#include <stdint.h>

Functions

int32_t	tiadalg_image_preprocessing_cn (void in_img[], int32_t in_width, int32_t in_height, int32_t in_stride, int32_t data_type, int32_t color_conv_type, float scale_val[], float mean_pixel[], int32_t pad_pixel[4], void out_img)
	It does basic image processing like plane deinterleaving, mean subtraction, scaling and padding. This API is intended to do image preprocessing for images beore feeding it to Deep Learning networks. More...

int32_t	tiadalg_image_preprocessing_c66 (void in_img[], int32_t in_width, int32_t in_height, int32_t in_pitch, int32_t data_type, int32_t color_conv_type, float scale_val, float mean_pixel, int32_t pad_pixel[4], void out_img)
	It does basic image processing like plane deinterleaving, mean subtraction, scaling and padding. This API is intended to do image preprocessing for images beore feeding it to Deep Learning networks. More...

Function Documentation

◆ tiadalg_image_preprocessing_cn()

int32_t tiadalg_image_preprocessing_cn	(	void *	in_img[],
		int32_t	in_width,
		int32_t	in_height,
		int32_t	in_stride,
		int32_t	data_type,
		int32_t	color_conv_type,
		float	scale_val[],
		float	mean_pixel[],
		int32_t	pad_pixel[4],
		void *	out_img
	)

It does basic image processing like plane deinterleaving, mean subtraction, scaling and padding. This API is intended to do image preprocessing for images beore feeding it to Deep Learning networks.

tiadalg_image_preprocessing_cn

Parameters

in_img	[in] : Input Image buffer
in_width	[in] : Input image width
in_height	[in] : Input image height
in_stride	[in] : Input image stride
data_type	[in] : data type of output buffer. Input data type is always assumed as U08
color_conv_type	[in] : input to output color conversion flag
scale_val	[in] : Scale value for each output plane pixels
mean_pixel	[in] : Mean value to be subtracted from each output plane pixels
pad_pixel	[in] : Number of pixel to be padded for output plane in left, top, right and bottom sides.
out_img	[out] : Output buffer, where data is generated continiously in memory

Note

Currently only 8 bit data type is supported
if Pixel subtraction values (mean_pixel) shoould be given in output plane order.
if Pixel scaling values (scale_val) shoould be given in output plane order.
Processing happens in floating point, and finally data is kept in U08/U16/S8/S16 container
For U16/S16 output buffer data is clipped according U8/S8 to match optmized flow, where processing happens alwaysin 8 bit domain.

◆ tiadalg_image_preprocessing_c66()

int32_t tiadalg_image_preprocessing_c66	(	void *	in_img[],
		int32_t	in_width,
		int32_t	in_height,
		int32_t	in_pitch,
		int32_t	data_type,
		int32_t	color_conv_type,
		float *	scale_val,
		float *	mean_pixel,
		int32_t	pad_pixel[4],
		void *	out_img
	)