Histogram computation on CUDA

Histogram computation is a frequently utilized task in image and video processing. As soon as histogram algorithm could be run in parallel, one can do that on GPU to get very high performance. We've implemented CUDA-based histogram kernels and now they are the part of Fastvideo Image & Video Processing SDK.

Applications for CUDA histogram

• Camera applications
• Digital Cinema
• High speed imaging
• Raw image processing

Features

• Input images: 8/12/16-bits per channel images in RGB, PGM, PPM, BMP, YUV, JPG, JP2 formats, byte array in CPU or GPU memory
• Input video: MJPEG (AVI), MJPEG2000
• Histograms: 256-bin, 512-bin, 1024-bin

OS Supported (64-bit)

• Windows-10/11
• Linux Ubuntu, L4T