Sobel Operation

Jalwinder Singh

Other Parts Discussed in Thread: CCSTUDIO

Hello,

I was trying to implement Sobel Operator from IMage library on DM6437. Since it was not working on Video Preview example, Bernie mentioned that the size of the input is YCbCr(16 bit) which is in-compatible with Image Library sobel function. There are many forms of sobel functions are available in image library which are as follows:

IMG_sobel_3x3_16s 3x3 Sobel Edge Detection for 16-bit Input,

IMG_sobel_5x5_16s 5x5 Sobel Edge Detection for 16-bit Input and some others,

Since my input data is 16 bit Ycbcr then why any of these sobel functions is not producing the sobel output on the display(as these are for 16 bit input)?

Also, library does not mention anything about output image pixel size.

Please provide any help..

thanks

Jalwinder

over 16 years ago

0 Jalwinder Singh over 16 years ago

Intellectual 280 points

In addition, are these sobel library functions compatible with 16-bit ycbcr input? if not, then how the sobel will work and on which input format?

thanks

Jalwinder

0 Gagan Maur over 16 years ago in reply to Jalwinder Singh

TI__Expert 8150 points

The functions are provided to work on planar data. You can either convert ycbcr data to planar data and use the provided functions or modify the provided code to perform sobel operation on ycbcr input. The source of the kernel is provided here: C:\CCStudio_v3.3\c64plus\imglib_v201\kernels\

Gagan

0 Jalwinder Singh over 16 years ago in reply to Gagan Maur

Intellectual 280 points

Hello Gagan and thanks for reply,

As you mentioned that YCbCr has to be converted into Planar data, Can you please mention in detail which planar data I need for Sobel operation and Is there any function available to convert YCbCr to planar.

thanks

Jalwinder

0 Bernie Thompson TI over 16 years ago in reply to Jalwinder Singh

TI__Mastermind 41665 points

Planar data means that you have individual buffers for each color component of the image, the drivers by default are giving you interleaved data in the form of a long array of bytes of Y Cb Y Cr Y Cb Y Cr, the planar version of this image would be three arrays one with Y Y Y Y Y Y, one with Cb Cb Cb, and one with Cr Cr Cr. To get these planar buffers you could use the code from the accessing pixels in a frame example as a starting point, and instead of just modifying values, setup the loop so that it copies each of the color components into individual arrays.

0 Jalwinder Singh over 16 years ago in reply to Bernie Thompson TI

Intellectual 280 points

Thanks for feedback Bernie.

Appreciated that one..

Jalwinder

Processors

Processors forum

Sobel Operation