Image Quality Assessment (IQA) 1.1.2
A fast, accurate, and reliable C library for measuring image quality.
Data Structures | Functions
include/iqa.h File Reference
#include "iqa_os.h"

Go to the source code of this file.

Data Structures

struct  iqa_ssim_args
struct  iqa_ms_ssim_args

Functions

float iqa_mse (const unsigned char *ref, const unsigned char *cmp, int w, int h, int stride)
float iqa_psnr (const unsigned char *ref, const unsigned char *cmp, int w, int h, int stride)
float iqa_ssim (const unsigned char *ref, const unsigned char *cmp, int w, int h, int stride, int gaussian, const struct iqa_ssim_args *args)
float iqa_ms_ssim (const unsigned char *ref, const unsigned char *cmp, int w, int h, int stride, const struct iqa_ms_ssim_args *args)

Function Documentation

float iqa_ms_ssim ( const unsigned char *  ref,
const unsigned char *  cmp,
int  w,
int  h,
int  stride,
const struct iqa_ms_ssim_args args 
)

Calculates the Multi-Scale Structural SIMilarity between 2 equal-sized 8-bit images. The default algorithm is MS-SSIM* proposed by Rouse/Hemami 2008.

See https://ece.uwaterloo.ca/~z70wang/publications/msssim.pdf and http://foulard.ece.cornell.edu/publications/dmr_hvei2008_paper.pdf

Note:
1. The images must have the same width, height, and stride.
2. The minimum image width or height is 2^(scales-1) * filter, where 'filter' is 11 if a Gaussian window is being used, or 9 otherwise.
Parameters:
refOriginal reference image
cmpDistorted image
wWidth of the images.
hHeight of the images.
strideThe length (in bytes) of each horizontal line in the image. This may be different from the image width.
argsOptional MS-SSIM arguments for fine control of the algorithm. 0 for defaults. Defaults are wang=0, scales=5, gaussian=1.
Returns:
The mean MS-SSIM over the entire image, or INFINITY if error.

Definition at line 129 of file ms_ssim.c.

float iqa_mse ( const unsigned char *  ref,
const unsigned char *  cmp,
int  w,
int  h,
int  stride 
)

Calculates the Mean Squared Error between 2 equal-sized 8-bit images.

Note:
The images must have the same width, height, and stride.
Parameters:
refOriginal reference image
cmpDistorted image
wWidth of the images
hHeight of the images
strideThe length (in bytes) of each horizontal line in the image. This may be different from the image width.
Returns:
The MSE.

Definition at line 37 of file mse.c.

float iqa_psnr ( const unsigned char *  ref,
const unsigned char *  cmp,
int  w,
int  h,
int  stride 
)

Calculates the Peak Signal-to-Noise-Ratio between 2 equal-sized 8-bit images.

Note:
The images must have the same width, height, and stride.
Parameters:
refOriginal reference image
cmpDistorted image
wWidth of the images
hHeight of the images
strideThe length (in bytes) of each horizontal line in the image. This may be different from the image width.
Returns:
The PSNR.

Definition at line 38 of file psnr.c.

float iqa_ssim ( const unsigned char *  ref,
const unsigned char *  cmp,
int  w,
int  h,
int  stride,
int  gaussian,
const struct iqa_ssim_args args 
)

Calculates the Structural SIMilarity between 2 equal-sized 8-bit images.

See https://ece.uwaterloo.ca/~z70wang/publications/ssim.html

Note:
The images must have the same width, height, and stride.
Parameters:
refOriginal reference image
cmpDistorted image
wWidth of the images
hHeight of the images
strideThe length (in bytes) of each horizontal line in the image. This may be different from the image width.
gaussian0 = 8x8 square window, 1 = 11x11 circular-symmetric Gaussian weighting.
argsOptional SSIM arguments for fine control of the algorithm. 0 for defaults. Defaults are a=b=g=1.0, L=255, K1=0.01, K2=0.03
Returns:
The mean SSIM over the entire image (MSSIM), or INFINITY if error.

Definition at line 59 of file ssim.c.

 All Data Structures Files Functions Variables Typedefs Defines


Copyright (c) 2011 by Tom Distler