How to use rfft2d_fr16 or rfft2d_fr32 on grayscale images

Hi Santhakumari,

I worked on rfft2d_fr16() , to validate this I used it's FFT output as input to ifft2d_fr16(), for FFT size 64 , twiddle stride 1, block exponent 12, scale 1 , but I am getting 12 bits scaled while compared input see below

1) My input is img[64*64 ]={0x7fff,0x7fff,0x7fff,0x7fff,0x7fff,.......0x7fff}

ifft2d_fr16 output is ={0x7,0x7,0x7,0x7,.......0x7};

2) My input is img[64*64 ]={0x4000,0x4000,0x4000,0x4000,........0x4000}

Ifft2d_fr16 output is ={0x4,0x4,0x4,0x4,.......0x4};

I changed my scaling values in both functions many times as well block exponent values also, but there is no change in ifft out.

If changed FFT_SIZE to 32 or 16 then output occuring like 8, 6 bit scaling respectively, my question is while I am taking 64*64 input image, should i use only 64 FFT size or other FFT size.

Another question is can i change my scaling factor .

Thanks and regards,

Santhosh 

Hi Santhakumari,

I worked on rfft2d_fr16() , to validate this I used it's FFT output as input to ifft2d_fr16(), for FFT size 64 , twiddle stride 1, block exponent 12, scale 1 , but I am getting 12 bits scaled while compared input see below

1) My input is img[64*64 ]={0x7fff,0x7fff,0x7fff,0x7fff,0x7fff,.......0x7fff}

ifft2d_fr16 output is ={0x7,0x7,0x7,0x7,.......0x7};

2) My input is img[64*64 ]={0x4000,0x4000,0x4000,0x4000,........0x4000}

Ifft2d_fr16 output is ={0x4,0x4,0x4,0x4,.......0x4};

I changed my scaling values in both functions many times as well block exponent values also, but there is no change in ifft out.

If changed FFT_SIZE to 32 or 16 then output occuring like 8, 6 bit scaling respectively, my question is while I am taking 64*64 input image, should i use only 64 FFT size or other FFT size.

Another question is can i change my scaling factor .

Thanks and regards,

Santhosh