Pay attention you can use different statistical parameters for different aims on image processing : for probabilistic description and classification of (different parts of) the images and for their quality estimation.

On one hand, you can use statistical parameters to caracterize the content of an image, its texture. Statistical methods can be further classified into first-order (one pixel), second-order (two pixels) and higher-order (three or more pixels) statistics . The basic difference is that first-order statistics estimate properties (e.g. average and

variance) of individual pixel values, ignoring the spatial interaction between image pixels, whereas second- and higher order statistics estimate properties of two or more pixel values occurring at specific locations relative to each other. So, first order measures are statistics calculated from the original image values, like variance, and do not consider pixel neighborhood relationships. Histogram based approach is based on

the intensity value concentrations on all or part of an image represented as a histogram. Common features include moments such as mean, variance, dispersion, mean square

value or average energy, entropy, skewness and kurtosis. Images can also be represented with high-order statistical parameters computed from co-occurrence or run-length matrices or from frequential approaches.

On another hand you can evaluate the quality of an image (after a filtering, a compression, etc.), with or without reference or with a reduced reference. PSNR, MSE, SSIM are quality metrics with reference (or fidelity metrics). but in recent state of the art there is a lot of quality metrics.

For more details on the two aspects of statistical parameters for image processing, you can read : Christine Fernandez-Maloigne, (Ed.)

Advanced Color Image Processing and Analysis

SPRINGER, Signals&Communications, 489 p., July 2012, ISBN 978-1-4419-6189-1

Some of the statistical parameters could be: First, second and higher order moments. Signal to Noise ratio, Integral Square error (ISE), Pearson statistics, etc.

Pay attention you can use different statistical parameters for different aims on image processing : for probabilistic description and classification of (different parts of) the images and for their quality estimation.

On one hand, you can use statistical parameters to caracterize the content of an image, its texture. Statistical methods can be further classified into first-order (one pixel), second-order (two pixels) and higher-order (three or more pixels) statistics . The basic difference is that first-order statistics estimate properties (e.g. average and

variance) of individual pixel values, ignoring the spatial interaction between image pixels, whereas second- and higher order statistics estimate properties of two or more pixel values occurring at specific locations relative to each other. So, first order measures are statistics calculated from the original image values, like variance, and do not consider pixel neighborhood relationships. Histogram based approach is based on

the intensity value concentrations on all or part of an image represented as a histogram. Common features include moments such as mean, variance, dispersion, mean square

value or average energy, entropy, skewness and kurtosis. Images can also be represented with high-order statistical parameters computed from co-occurrence or run-length matrices or from frequential approaches.

On another hand you can evaluate the quality of an image (after a filtering, a compression, etc.), with or without reference or with a reduced reference. PSNR, MSE, SSIM are quality metrics with reference (or fidelity metrics). but in recent state of the art there is a lot of quality metrics.

For more details on the two aspects of statistical parameters for image processing, you can read : Christine Fernandez-Maloigne, (Ed.)

Advanced Color Image Processing and Analysis

SPRINGER, Signals&Communications, 489 p., July 2012, ISBN 978-1-4419-6189-1

Statistical parameters are "parameters or indexes" for probability distributions, like the the mean and variance for a normal distribution?

So may be you are asking for quality descriptors or "statistical descriptors"?

Statistics are simple tool that help us for better understanding of our images. Which characteristics do you need to measure? If you need to measure noise you need use "dispersion statistics" based in variance; if you need to measure mean signal you need use "central statistics".

sse, msereg etc are used for finding the error

Image  quality assessment problem can be categorized into 3 following  classes

1) Full reference image quality assessment (FR-IQA)

2) Reduced reference image quality assessment (RR-IQA)

3) No reference image quality assessment (NR-IQA)

For FR-IQA usually used: Structural Similarity Index (SSIM), Fast SSIM  and Peak signal to noise ration (PSNR)

For RR-IQA use a training approach to evaluate the quality of an image.

For NR-IQA give the quality score by processing the test image. There are

different parameters can be used to evaluate the quality of no reference image.Some of the parameters used for this kind of images are Anisotropy, Discrete Cosine transform (DCT), wavelet and Gabor filtering. You can evaluate the quality of NR-IQA by: Blind Image Qualitymeasure (CBIQ), Learning based Blind

Image Quality measure (LBIQ), Distortion Identification-based

Image Integrity and Verity Evaluation (DIIVINE) index , Blind

Image Notator using Discrete cosine transform (DCT) Statistics

(BLINDS-II) 

Is it mean and covariance parameters enough to recognize the face? and I also want a simple matlab code for face recognition?please help me.....