Novel OpenGL Pixel Shader Dewarping
medium.com
Novel OpenGL Pixel Shader Dewarping
1–8 of 8 posts
Re: Novel OpenGL Pixel Shader Dewarping
#2Re: Novel OpenGL Pixel Shader Dewarping
#3You'll probably catch a bit of flak for pitching this as novel, but I really like how you've packaged it into all into a purpose built application. Those kinds of apps are the most fun to tinker with and feel rare these days :).
Re: Novel OpenGL Pixel Shader Dewarping
#4Results look great. Would be nice to have a performance comparison in numbers against other methods.
Re: Novel OpenGL Pixel Shader Dewarping
#5Not "novel", I was doing exactly this 10 years ago for a videoconferencing device.
Re: Novel OpenGL Pixel Shader Dewarping
#6From what I understand from code those unwarps are just doing matrix multiply to get unwraped pixel location? In this case doing these operations directly in fragment shader instead of texture lookup will be faster. Memory bandwidth is not free. But simple ALU like this (just couple FMA's) can easily hide in shadow of texture sampling that happens afterwards. So simply upload those undistortion matrices (mat1 & mat2) as uniforms and do matrix multiply in shader for adjusting texcoords.
Re: Novel OpenGL Pixel Shader Dewarping
#7From what I understand from code those unwarps are just doing matrix multiply to get unwraped pixel location? In this case doing these operations directly in fragment shader instead of texture lookup will be faster. Memory bandwidth is not free. But simple ALU like this (just couple FMA's) can easily hide in shadow of texture sampling that happens afterwards. So simply upload those undistortion matrices (mat1 & mat2)…
map1 and map2 are the same dimensions as the video image.
Re: Novel OpenGL Pixel Shader Dewarping
#8From what I understand from code those unwarps are just doing matrix multiply to get unwraped pixel location? In this case doing these operations directly in fragment shader instead of texture lookup will be faster. Memory bandwidth is not free. But simple ALU like this (just couple FMA's) can easily hide in shadow of texture sampling that happens afterwards. So simply upload those undistortion matrices (mat1 & mat2)…
map1 and map2 are the same dimensions as the video image.
Ah, I see. Yeah, then the current approach is fine.