This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Another question about DM368IPNC-MT5 Reference Design Camera

This question is slightly off topic, but I  think there could be an answer here....

The DM368IPNC camera employed a Sony 5M pixels CMOS sensor, The Bayer pattern will be used to filter the color while the camera samples the input images.

The Bayer's color filter arrays mosaic forms by 1 red filter, 1 blue filter and 2 green filters. The 5M pixel sensor is divided into 1.25M pixels groups for red and blue and 2.5M pixels for green.

Does it mean the native resolution for this camera is 1280x720 = 921,600 pixels (< 1.25 Mpixels) ?

BTW, I kept sending the request to talk with Appropho's engineers, the request has not yet fulfilled. A moment ago,  I tried to get on their website :http://www.appropho.com/NewWeb/Product_DM368MT5_Spec.php#a

It is not accessible today ... hope it is just the web traffic's issue.

  • Kien Nguyen said:

    The Bayer's color filter arrays mosaic forms by 1 red filter, 1 blue filter and 2 green filters. The 5M pixel sensor is divided into 1.25M pixels groups for red and blue and 2.5M pixels for green.

    Does it mean the native resolution for this camera is 1280x720 = 921,600 pixels (< 1.25 Mpixels) ?

    When you demosaic the Bayer pattern image you calculate the missing two colors for each pixel based on the surrounding pixels, thus you keep the same resolution on your resulting output image as the sensor has for its Bayer pattern. In other words the 5MP image sensor generates an actual 5MP output image after Bayer demosaicing, each red, green, and blue filtered pixel becomes a full color pixel, interpolated based on its neighbors. The actual native resolution of the sensor/camera would then depend on the aspect ratio of the sensor, which I have not been able to quickly find without the sensor part number or datasheet.

    Kien Nguyen said:

    BTW, I kept sending the request to talk with Appropho's engineers, the request has not yet fulfilled. A moment ago,  I tried to get on their website :http://www.appropho.com/NewWeb/Product_DM368MT5_Spec.php#a

    It is not accessible today ... hope it is just the web traffic's issue.

    I just tried that link, and it opened for me, so the site may just be having intermittent issues.

     

  • Hi Kien,

    Kien Nguyen said:

    The DM368IPNC camera employed a Sony 5M pixels CMOS sensor, The Bayer pattern will be used to filter the color while the camera samples the input images.

    The Bayer's color filter arrays mosaic forms by 1 red filter, 1 blue filter and 2 green filters. The 5M pixel sensor is divided into 1.25M pixels groups for red and blue and 2.5M pixels for green.

    Does it mean the native resolution for this camera is 1280x720 = 921,600 pixels (< 1.25 Mpixels) ?

    Firstly, i think you had a typo as we support 5MP Aptina CMOS sensor and not Sony. Let me know if it is mentioned on some TI or Appro website.

    Anyways, coming to the Bayer Pattern question. Technically, you are right in the understanding of the bayer pattern. There is a Green, Red or Blue filter on each pixel that responds to the intensity of light. Now this means based on the pattern like RGGB, you would have 2 green filters and 1 Red and 1 Blue filter. Now for example, if we have a read at a particular pixel, but no green or blue, we need all colors to get the full information of that pixel. These 'missing' values are recovered using CFA interpolation. It is explained in more details on the Programmers Guide of VPFE or can be read on any Wiki page.

    Kien Nguyen said:

    BTW, I kept sending the request to talk with Appropho's engineers, the request has not yet fulfilled. A moment ago,  I tried to get on their website :http://www.appropho.com/NewWeb/Product_DM368MT5_Spec.php#a

    It is not accessible today ... hope it is just the web traffic's issue.

    I checked the URL at Appro side and it seems to be working at this time. I guess, it would have been some traffic or intermittent problem with their web server.

    Let us know if you face issues in interacting or getting support on your IPNC Reference Design from Appro.

    Regards,

    Anshuman

  • Anshuman Saxena said:

    Hi Kien,

    The DM368IPNC camera employed a Sony 5M pixels CMOS sensor, The Bayer pattern will be used to filter the color while the camera samples the input images.

    The Bayer's color filter arrays mosaic forms by 1 red filter, 1 blue filter and 2 green filters. The 5M pixel sensor is divided into 1.25M pixels groups for red and blue and 2.5M pixels for green.

    Does it mean the native resolution for this camera is 1280x720 = 921,600 pixels (< 1.25 Mpixels) ?

     

    Firstly, i think you had a typo as we support 5MP Aptina CMOS sensor and not Sony. Let me know if it is mentioned on some TI or Appro website.

    Anyways, coming to the Bayer Pattern question. Technically, you are right in the understanding of the bayer pattern. There is a Green, Red or Blue filter on each pixel that responds to the intensity of light. Now this means based on the pattern like RGGB, you would have 2 green filters and 1 Red and 1 Blue filter. Now for example, if we have a read at a particular pixel, but no green or blue, we need all colors to get the full information of that pixel. These 'missing' values are recovered using CFA interpolation. It is explained in more details on the Programmers Guide of VPFE or can be read on any Wiki page.

    I checked the URL at Appro side and it seems to be working at this time. I guess, it would have been some traffic or intermittent problem with their web server.

    Let us know if you face issues in interacting or getting support on your IPNC Reference Design from Appro.

    Regards,

    Anshuman

    [/quote]

    Hi Anshuman,

    Yes, it is my mistake,  the sensor is 5MP Aptina CMOS. I just verified the sensor posted in Appropho's website.

    Appropho has not given the response regarding the implementation of 1280x720P60 feature, perhaps they are waiting for the commitment from TI ?

    I am still waiting for TI engineer/sales contact me regards other question (Service Request#1-642741713)

     

    Regards,

     

    Kien

  • Bernie Thompson said:

    The Bayer's color filter arrays mosaic forms by 1 red filter, 1 blue filter and 2 green filters. The 5M pixel sensor is divided into 1.25M pixels groups for red and blue and 2.5M pixels for green.

    Does it mean the native resolution for this camera is 1280x720 = 921,600 pixels (< 1.25 Mpixels) ?

    When you demosaic the Bayer pattern image you calculate the missing two colors for each pixel based on the surrounding pixels, thus you keep the same resolution on your resulting output image as the sensor has for its Bayer pattern. In other words the 5MP image sensor generates an actual 5MP output image after Bayer demosaicing, each red, green, and blue filtered pixel becomes a full color pixel, interpolated based on its neighbors. The actual native resolution of the sensor/camera would then depend on the aspect ratio of the sensor, which I have not been able to quickly find without the sensor part number or datasheet.

    [/quote]

    When you find out about the aspect ratio of that 5MP Aptina CMOS sensor and the camera native resolution, please let me know.

    Thank Bernie.

    Regards,

    Kien

  • Kien Nguyen said:
    When you find out about the aspect ratio of that 5MP Aptina CMOS sensor and the camera native resolution, please let me know.

    I took a deeper look and saw that MT9P031 is shown on their block diagram, checking the Aptina site shows that the sensor is likely a MT9P031I12STC, which has a 4:3 aspect ratio and a maximum resolution of 2,592H x 1,944V making up the 5MP image. Note that you may have to ask Appro about what resolutions their sensor driver actually supports, their collateral implies that it only supports up to 1920x1080, so you may not be able to effectively use the full resolution of the sensor with their existing driver software.

  • Bernie, Kien,

    Current IP Camera Reference Design software that you have might not have support for more than 1920x1080. But we have an internal development version where we do a 5MP capture from the same sensor and encode it using JPEG.

    In the later releases, we will add support for 5MP capture in IP Camera for external customers too.

    Regards,

    Anshuman

  • Anshuman Saxena said:

    Bernie, Kien,

    Current IP Camera Reference Design software that you have might not have support for more than 1920x1080. But we have an internal development version where we do a 5MP capture from the same sensor and encode it using JPEG.

    In the later releases, we will add support for 5MP capture in IP Camera for external customers too.

    Regards,

    Anshuman

     

    Great news, but I am working on a "broadcasting-type" video so the resolution I deal with in the near future (10 years) is mostly 1920x1080. Video/imaging folks in Panasonic, JVC etc..  are talking about 4k  televisions. I can't imagine this will be in the consumer 's living room soon.

    Imager and encoder can easily handle H264 1280x720P60 and 1920x1080i60 is what I am looking for.

    BTW, thank you both for clear up my misunderstanding about the effect of the Bayer pattern on image sensor. Let me test my understanding, in theory this camera sensor can produce an image resolution up to 2592x1944 (5MP). However, TI and Appropho current drivers only support the resolution up to 1920x1080  due to various constrainsts, limitations of the current design ?

    Regards,

     

    Kien