EmbeddedRelated.com
Forums

MT9D111 camera defected pixels problem

Started by sysizlayan 6 years ago2 replieslatest reply 6 years ago113 views

Hi,
I am currently working on an IoT project involving cameras.

One of the cameras we are using is Aptina-MT9D111 which is controlled by a STM32 processor.

The camera is turned on, a snapshot is taken then it is turned off again.

However, the first 8x512 pixel area from the top left corner of the image is similar to rest of the image but corrupted. Similar thing happens at the last 8x1200 pixel area at the bottom.

Example image is given in the link below:
http://tinypic.com/r/rbk36o/9

The camera just gives the JPEG decodec image from DCMI and we are adding JPEG header and footer to the camera. We are expecting the corruption occurs because of the header/footer image mismatch.

Do you have any comments on this behavior of the camera?  


[ - ]
Reply by rtomkinsMay 3, 2019

I see the upper and lower anomalies in the image that you speak of. I found the data sheet for this camera here.

Page 36, says, The sensor core pixel array is configured as 1,688 columns by 1,256 rows (shown in Figure 14). The first 52 columns and the first and the last 20 rows of pixels are optically black and are used for the automatic black level adjustment. The last 4 columns are also optically black.

Have you properly accounted for this in your code?

There are a number of different output formats with this device as well as a number different output interfaces available.

Possibly, your code may not be handily accommodating the choices you made as to the output interface or the output format.

You should probably elaborate more on the attributes you have set, and the interface you are using.

[ - ]
Reply by mr_banditMay 3, 2019

What rtomkins said.

Read the datasheet. Not once or twice. It can easily take 5..10 passes through it before you really get a grasp on it.

Go thru the register setup && grok every setting. Look at the default settings. I suspect you are using the defaults. 

I cannot look at the image - corporate firewall. 

HOwever, you have an important first step - you have an image. Not trivial. SUccess to you.