This section provides a description of the functions that are applicable to image filtering and format conversions.
Color space conversion from YCbCr to RGB enables the display of digital video data generated, for instance, by an MPEG or JPEG decoder system on RGB displays.
These routines take a packed YCrYCb color buffer in big-endian or little-endian format and expand the constituent color elements into separate buffers in little-endian byte ordering.
The convolution functions are used to apply generic filters to the input image. Filter sizes of 3x3, 5x5, 7x7, and 11x11 are supported. Typical applications include, but are not limited to, image smoothing and sharpening.
Correlation functions are provided to enable image matching. Image matching is useful in applications such as machine vision, medical imaging, and security/defense. The following functions implement highly optimized corrrelation for commonly-used filter sizes such as 3x3, 5x5, and 11x11.
The functions below are more generic and can implement correlation for user-specified pixel neighborhood dimensions within documented constraints. The IMG_corr_gen_iq function handles 32-bit Q-point data.
Error diffusion with binary valued output is useful in printing applications. The most widely-used error diffusion algorithm is the Floyd-Steinberg algorithm. These functions are optimized implementations of this algorithm.
Median filtering is used in image restoration, to minimize the effects of impulsive noise in imagery. Applications can cover almost any area where impulsive noise may be a problem, including security/defense, machine vision, and video compression systems. Optimized implementation of the median filter for 3x3 pixel neighborhood is provided in the routines below.
The routines IMG_pix_expand and IMG_pix_sat, respectively, expand 8-bit pixels to 16-bit quantities via zero extension, and saturate 16-bit signed numbers to 8-bit unsigned numbers. They can be used to prepare input and output data for other routines such as the horizontal and vertical scaling routines.
Copyright 2012, Texas Instruments Incorporated