Week 2 Exercise: Experiment With Camera Compression

i use the image in the link and use the snapshot with the default size ,save under JPEG format the result was 64KB ,
than resize zooming the photo Multiplier and take the snapshot ,save under same format the result was 124KB
it is right
but how calculate the bandwidth ??

it seem the second snapshot (124kB0) better

GV-FD320D 3MP Camera

Hi-Res with 3MP camera at max is 2048 x 1536 with size of 410KB, bandwidth at 76Mbps, after lowering compression, Low-Res is at 1024 x 768 with size of 141KB, bandwith at 20Mbps.

AXIS model #m3114 running at .20 Mbit/s at 1024 X768 resolution 10fps and compression set to 30.

With compression moved to 80 went down to between .06 and .08 Mbit/s

Hi res picture of my youth running team versus an edited one.

I used a camera feed from an Axis 206 (which I believe has been discontinued).

I started with resolution set at 640X480 and Compression level at "Low" (setting available). The snapshot was 28.3kb.

I changed the camera compression to "High". The image was noticeably artifacted but I could still recognize objects. The resulting snapshot was reduced to 9.29kb...1/3 less size than the original. This image could be acceptable depending on what level of detail you desired to see.

I used an Axis M3005-V set to 1920x1080. I started with the default compression level of 30, raised the level to 60, and then 90.

Compression = Default 30 - Clear picture - file size of 177KB

Compression = 60 - slight bluring in fine lines of painting - file size of 106KB

Compression = 90 - A lot of artifacting (pixelization) - file size of 47.5KB

I did the exercise using the provided image, exported with various quality settings from GIMP 2. The original image was 61 KB on my disk.

Exported at quality level 45:

This occupies 40K on the drive, and I can still read the number on the man in the lower left.

Exported at quality level 25:

This occupies 26K, and I am left guessing about that number.

Exported at quality level 5:

This occupies 11K, and does not seem useful for much.

I took a snapshot from a Axis M3014 at 1024 x 640, and it was 57KB. I changed to 640 x 480 and the snapshot was 29 KB. There was a noticable difference between the two images. Below is a comparison showing a zoomed in portion.

Using the pix in the link, we get 500 pixels wide by 333 pixels vertically with file size 60.5kb. Using Windows Photo Viewer, it can be zoomed/expanded (compressed?) about 400% and still read numbers on runners. Using Paint, the zoom/expansion of about 300+% still provided reasonable detail. May not have the best photo-editing software for the exercise.

This is the initial image and its about 64kb on disk

@50% compression its size on disk is about 36kb

while at 80% compression its 24kb.

H.264 - 24kb MJPEG - 33Kb

Using the pix in the link, we get 500 pixels wide by 333 pixels vertically with file size 60.5kb. zoomed/expanded using Paint, 768x614 pixels size 210kb with reasonable detail but worseimages.

H264 (128KB) excellent quality

H264 (104KB) lowest quality

Using the image in the link (saving with Irfanview and adjusting quality) I tried 3 settings in here.

Save image, no adjustments, 62KB

Quality cranked to 100, 144KB (almost un-noticable picture inprovement)

Quality set to 10, 13.2KB (very low quality, pixelation, colors almost seem bland)

Considering these results, in a security environment, when saving say a single Jpeg and size isn't really a concern for a single snapshot (144KB is fairly small and easily email-able) I would think it would be best to save as large as possible to get the most clarity? I know the implications of larger images for video but it would make sense to me to provide a single image in the highest quality possible. Any thoughts on that guys?

EDIT: I am aware that the date is one day ahead. These cameras are mine and were installed a week ago. Just fixed/synced time. These cameras were simply plugged in and set to record.

I learned quite a bit about the Dahua IPC-HDBW4300E. I learned that Dahua doesn't really want me to compress that much. I forced the camera to black and white to try and get the best comparison, but the camera doesn't really want me to compress that much. In color the but rate remains high. I can reduce the bandwidth quote a bit in color only if I set it to manual exposure because the gain is too high in auto. At quality 6, the DRF is at 14 with a max allowable bitrate of 8261kb/s. At Quality 1, the DRF is at 34 with an average of 731kb/s. I could set the threshold to a minimum of 768kb/s but I left both at the max threshold in VBR to get a fair comparison. The framerate was set to 10fps.

I found that it really doesn't look that bad at Q1 and saves on bandwidth. We found the same thing after analyzing at a jobsite today and we set all at Q1 with a 4Mb threshold cap. This is much different than other manufactures. I may try this on some Hikvision cameras next.

Q1 3MP:

Q6 3MP:

This was a high motion scene as the leasves are moving and falling. I wanted to try a couple things with this set of photos. The first photo is H.264 CBR @ 3072 k/bits. I wanted to see how much difference it would make by going to VBR with same settings. Compression rate was set low for the VBR example. In the 3rd photo I used high compression.

CBR 2.8Mbits

VBR low compression 1.9Mbits

VBR Higher compression 1.6Mbits

As you can see the amount of bandwidth varied greatly between the CBR and VBR. 32% reduction.

With the higher compression setting using VBR gained approx 16% over the 2nd image and 43% ove the CBR image. The 3rd image has the worst quality but there is not much difference from image 1 and 2 to warrant using CBR. I was suspecting a higher badwidth drop >16% from image 2 to 3 but the image had a great deal of motion and my VBR cap was set to 3072 Kbits. ????

I plan on lower the VBR cap and see what that yields

Tim

The image file is 60.5KB, resolution 500x333 pixels, compression value is not indicated and cannot be selected to be changed; how could I change the compression value to compare the quality difference?

Was having difficulty finding a program to change the quality, so i took pics with my phone at 8meg and vga. Below are the side by side images. I also attached a screen shot showing the file resolutions of the 4 pics i took (8mg,3mg,1mg, vga).

After exporting original into a JPEG file using GIMP and changing quality settings to 20 it changed the bandwidth to 25.3kB ...resulting in a savings of over 70% bandwidth over original 92.2 kb ...with the same 500x334 resolution.

Below is the original above is after changing the quality settings from 90 to 20.

I had a hard time resovling this ,the picture in the file you gave us .

can you help on this?I understand if you raise the compression level the image

will not be as sharp.also reducing band width

Axis M3005 at 800X600 and GOV length of 30 and at 10fps.

Adjusted to 1280X960, GOV to 75 and still 10fps.

The one interesting thing about the M3005 is the constant bit rate with a cap of 10000kb/s streams the video at an actual 10000kb/s, instead of doing it at a reasonable rate to give a good image, but not to exceed the set amount. Other Axis cameras will stream at a reasonable rate with the cap, but this one automatically streams at the cap set. Weird

I was having some issues streaming video through VLC so I used the photo that was provided.

Original Photo is 61 KB. I zoomed in on each photo so you can see the change clearly.

Second photo is exported at 50 % quality. File size is 42 KB.

Third photo is exported at 10% quality. File size is 17KB.

I did not have a live stream available so I used this image:

https://ipvm-uploads.s3.amazonaws.com/uploads/4d09/254b/NYC-Marathon%20(1).jpg

Since I can't do anything with compression, I resided the image to see different spoecs.

Here are the before and after specs after I resized the image smaller using Paint.net

Original Image:

NYC-Marathon%20(1).jpg

File Size 62.0 kB

Pixels 500x333

Number of pixels: 166,500

Print Size 6.944 x 4.625 inches

Resolution 72 x 72 ppi

Resized Image:

NYC-Marathon%20(2).jpg

File Size 24.9 kB

Pixels 212x141

Number of pixels: 28,892

Print Size 2.208x1.469 inches

Resolution 96x96 ppi

I performed this exersice with an Eagle Eye Networks camera, model number EN-CDUM-002. This manufacturer offered three settings(labeled as quality) high, medium and low. Images and measuremnts are depicted below:

High - 80.5kb

Medium - 45.9kb

Low - 22.4kb

Lot of good responses coming in. One thing I would encourage is to take snapshots when a person is in the field of view, ideally walking, which will better show maximum loss / degradation.

Used GIMP to export original 61KB to a .jpg file at different quality levels producing the following:

At 50 Q level: 42KB file--Image still clear--no discernible degradation

At 25 Q level: 30KB file--Image begins to degrade slightly--Numbers are fuzzy around edges

At 5 Q level: 11KB file--Image very degraded. Unable to read numbers on shirts

I opened the JPEG image (60.5 KB) from the site using PHOXO photo editor and then saved it in the same format with 50% quality (34.5 KB). I kept the remaining options the same.

the result:

1) the reduction of size was around 43%.

2) No visual effect was noticed.

.

Thanks

I took a few snapshots from a 3Mpix camera.

Available quality levels were 13-90, with 13 being the highest compression and 90 having the least compression.

I started at 80 and resulted an image size of 696KB.

I dropped quality to 60 and resulted in an image size of 454KB. There were virtually no visible artifacts unless digitally zoomed in a great deal.

I then dropped quality to 13 and resulted in an image size of 149KB. There were very visible compression artifacts.

It may be tough to see at first, but if you look at the roof in the last image you can see some artifacts.

I used a photo 1936X1288 Size745 KB and compressed 80x and had a result of a photo a photo 1936x 1288 Size 67.2. The photo was greatly decreased in size and the quality decreased quite a bit.

Used Avigilon ViQ-HD-CRD225-3-8

High Resolution Stream 1920 x 1080, 15fps, 4.5mbps

Low Resolution Stream 960 x 540, 5fps, 250 kbps

I used an AXIS P3365-v camera outdoor in the late afternoon just before sunset for the experiment. The effect of various compression levels in outdoor environment as well as moving objects (vehicles) was studied.

I couldnt find the rtsp url at Soleratec, and the AXIS website gave urls for players such as Windows Media Player and Quicktime. However I managed to find the rtsp url for VLC with some good guess and trial and error.

Now for the experiment, I had considered three different compression levels - 30%, 70% and 90%, while the CBR was set at 2000kbps. I had recorded the video clips in VLC, and then exported into AVInaptic for analysis. Though the tool was saying that "analysis is not complete", which I didnt understand what the cause could be or see parameters such as DRF, the bitrate and frame rate read as follows:

@ 30% compression, 870.3kbps; 17.45 fps

@ 70% compression, 266.16 kbps; 17.57 fps

@ 90% compression, 225.83 kbps, 17.54 fps

Evidently, the quality and bandwidth was the highest when the compression level was set at 30%. At 70%, the bandwidth was reduced but the quality levels dropped as well with some visible artifacts, but still tolerable. At 90%, the artifacts were worse and no visible change in bandwidth was noticeable.

I would like to send the VLC recorded files and the AVInaptic analysis results which I have saved as a text file, but have some difficulty in trying to attached in here as I couldnt locate an 'Upload' button, which was there for uploading the images though.

A few minutes of advisory class on how to attach/upload pictures, videos and text before start of todays class could be useful.

As you can (hopefully) see from the insert image is now ~42k file. Curious that this is non-linear? moving from the default 9 to 5 does not halve the file size. I recognize that JPEG is already a compressed format and this may have something to do with it?

Brian

I downloaded three images from one of the Axis P3304 cameras changing the "quality" settings from 1 to 5 to 10. I could see very little difference in the image doing that. I am thinking that "quality" in Axis was compression. Is that not correct?

Now there is a correlation of "quality" to average image size at 5fps;

Quality setting of 1 = AIS 7500bytes

Quality setting of 5 = AIS 11333bytes

Quality setting of 10 = AIS 36040bytes

The resolution of the camera is set to 1280X800. If I change the resolution of the camera, I'm sure that I will see a greater variation in the image.

Now if I look at m Avigilon 5.0-H3-DP1 at quality setting 6 I get 12.8 Mbps while at quality setting 20 I get 2.3Mbps.

Using this camera I can see the difference between the 2 views when zooming into the details.

This was my first time using program like Gimp but by reading the forum was able to demonstrate how compressing the image can still provide a decent clear image from 64kb disk space to 40kb or less disk space.

Work has been killing me lately, leading to my late submission (not complete) so i'm going to have to take down a camera at the job site i'm at tomorrow and bring it to my hotel to do this. I do have a question however, as I had a brief look at this.

Pulling a stream out of axis http url to vlc.... it was streaming at mjpeg, and the stream in vlc is washed out red and pixelly. Any idea's?

Also when i change the stream to H.264 the url (right click on the feed as it's an imbedded stream) it still only gave me the url for Mjpeg. Anyone had a similar problem? or is there something i'm doing wrong.

I'll have a play tomorrow night, with a different laptop as well just in case my work laptop is doing something swirly.

Hello
The GIMP software is not clear to lower the compression/quality of the provided image.
Therefore, I used the adobe photoshop element to change the quality of the provided image, I used the option Save For Web, fixed the image size for 500x333 pixels, and set the quality as 10%.
Below is the resulting image.

The resulting image size is 27 KB.

Dear Marie, John & Colleagues

For a better organized examination, below link of PDF file for your reference, let me know if you face any problems while downloading it.

I hope the method of calculating the BW I used is correct, the one we got in readings can only be applied if pixels are changed, fixing the bits value ( 8, 16, 24..) or FPS..., while in our case the pixel counts are fixed and the variable factor is the KB value..please correct me if I am wrong.

Thank you.
Regards,

AJ

i was working with the image provided on cours.

with Gimps i use difrent compresion

at 100% quality : size 500x333 disk size 218.77kb

at 65% quality inage is still clear, disk size 55.52kb

at 20%qualiti image is losing quality, disk size 25.21kb

i have to aplolagize for the late reply to this exercise but i was werry busy at work.

Camera: Axis P3354 Fixed dome Day/Night

Early morning shot, I wanted to see how the light would effect the bandwidth as I didn't have time or access to a 'busy' shot. These camera's also have Axis' billed 'lightfinder' technology so I wanted some average light in there to try and make use of this, and subsequently saw the following.

Resolution 1280x960

0 Compression File size: 1.56 Mb

30 Compression File Size: 970Kb

60 Compression File Size: 842 Kb

90 Compression File Size: 763Kb

Interesting to see that while there's a huge difference between the uncompressed file and the first compression, between 30-90 compression there is only a difference or saving or 207Kb. The picture quality however drops significantly past 30 compression which is the axis default level. A few of our contracts are street surveillance for council's... which by the side incorporate alot of police evidence. We are yet to raise the axis compression level higher than 30. They (for the moment, mid-transition into ip upgrades) have the storage to support it so until their ip network grows too large and storage becomes an issue, I think we'll stick with this as it's a good mid-ground.

Greetings,

I used the marathon picture (originally 62 KB) provided. Without changing resolution (500 x 333 pixels), and only adjusting quality using Gimp , here are the picture results (all jpegs):

Image quality: lowered 50% (quality 45), File size: 40KB

Image quality: lowered 75% (quality 20), File size 26 KB

As the quality of the picture is lowered, the storage size is smaller. If this was video, the bandwidth used would also be lower. The picture quality does degrade from the original to compressing at 50% but the quality is good enough to clearly capture most runner's numbers and facial features. The 75% compression still has decent facial recognition but quality is getting very low. From a sales perspective, decent (but not great) quality will definitely reduce cost for the customer. Boom!!

Look at the shadow at the top of each pic.

I used an Axis P3304 camera at 720P
Default compression setting of 30 resulted in:
Bitrate: 1155 kbps
Average DRF: 29.87
File size for a 30 second clip at 15ips: 4.12 MB

Resulting image capture:

Higher compression setting of 80 resulted in:
Bitrate: 291.54 kbps
Average DRF: 38.80
File size for a 30 second clip at 15ips: 1.03 MB

Resulting image capture:

You can see some degradation in the lines but it becomes really noticable in the image of the person's face. However, for the tradeoff of bandwidth and storage, all else being equal, the difference is worth the slightly lesser quality.

Brian

Excellent point John, thank you.

I used the image in the link which was 60.5 KB, and compressed the image to 43.8 then 24.6. The quiality decreased of course, especially when I zoomed in on the compressed photos. How do I figure out the bandwidth for these images, or is that something I can only do if I take the screenshot from one of our cameras?

Thanks, sorry for being behind, work has been a bit hectic.

Whoops, I see your response to Emaldin about measuring bandwidth.

Here the images I compressed.

compressed to 43.8kb

compressed to 24.6kb

I have selected Printer testing table for this assignment

The original table (100% quality) is 800x533 and 175 KBytes of disk space

Now after reducing quality to 40% some noise was added to picture and its volume is now 44 KBytes. The chart takes only 25% of disk space and quality degradation is not too high so this conversion is beneficial.

And the third picture shows quality of 5% from original. Image now consume 19.9 KBytes of space but quality is too low with major artifacts and loss of minor colors.

I used the marathon picture for this exercise. Without changing any settings, the picture was 61kb:

After changing the quality levels to 50%, I got the following image which was 42kb.

There isn't much less detail in this image but there is a significant reduction in the size.

100% Quality 218 KB

70% Quality 58.7 KB

40% Quality 38.1 KB

One can see as the file size shrinks sharpness and the quality of the image decreases as the compression scheme merges similar colors into one creating larger and larger "grains."

I used GIMP to export the original image (61KB) to a .jpg file at different quality levels.

We can see the differences for 2 Quality level (I zoomed on the clouds in the sky):

At 50 Q level: 34KB

At 20 Q level: 21KB

Block artefacts are clearly visible when compression is applied.

John, I understand your point of course but for me this part of the image (cloud) is where the default are the more visible if I do not zoom out. Compression is usually more visible on large area of uniform colors than on area with details.

For the purpose of video surveillance obviously we are all more focus on people than on clouds!

I see noticeable difference between the two images in the facial details and the numbered tags on the runners...both of which would be key parts of the image in surveillance.