Test System for CMOS Digital Image Sensors

An optional background subtraction algorithm can be applied, where appropriate, to remove lens-shading effects. Enabling background subtraction, results in standard deviations that more closely relate to the amount of noise in each of the color planes. The image and response uniformity graph, shown at left, has background subtraction disabled and the lens shading fall-off is very visible.

Bayer Photo Response Uniformity Algorithm

The Image Sensor Lab Bayer Photo Response Uniformity Algorithm first calculates the mean and standard deviation of all the pixel intensities in the entire image, for each of the four Bayer color planes.

An optional background subtraction image processing step can be selected that can help eliminate lens shading effects, when using this test as a measure of the pixel noise in the image. The correction algorithm uses an Nth order filter and has user configurable kernel size and percentile threshold algorithm inputs.

The algorithm also calculates maximum neighbor non-uniformity across a user specified grid of cells. The results are returned as a maximum percentage deviation.

The individual Bayer color planes can be displayed as well as the percentage deviation calculations converted to an image display. 3D views of both the captured image pixel intensities as well as the percentage deviation data are also available.

A summary table is output that contains the results for each color plane. The mean intensity and standard deviation are reported, as is the maximum neighbor deviation.

Limits can be specified separately for each color plane. The intensity mean minimum and maximum, as well as the intensity standard deviation maximum, can be specified.

Input Image

A Bayer image type is required for this algorithm.

Background Subtraction On/Off

Specifies if the background subtraction algorithm is enabled.

Background Subtraction Filter Nth Order Kernel Size

Specifies the X and Y grid size of the kernel used for the Nth order filter.

Background Subtraction Filter Nth Order Percentile Threshold

Specifies the percent threshold used to order the classification of the filter.

Analysis Results Table

A results table, formatted as a tab-delimited string, is returned with a row of data for each Bayer color plane. The mean intensity and intensity standard deviation is reported.

Bayer plane image

A drop list allows the selection of which Bayer plane will be displayed in the image window. 3D view is also available.

Bayer Photo Response Non-Uniformity

This test measures photo response non-uniformity across the surface of the sensor. The lens and sensor have non-linear responses and most CMOS image sensors have obvious visible intensity fall-off in the outer corners. The non-uniformity test first divides the image into a grid of regions and then calculates the % intensity deviation with any neighbor region. The maximum neighbor deviation is then recorded and displayed for each defined region in the image.

The non-uniformity percentages are presented in a table and additionally in an image format.

Input Image

A Bayer image type is required for this algorithm.

Non-uniformity Grid X cells

Specifies the horizontal grid size (in pixels) used for defining regions of interest used in the non-uniformity calculation.

Non-uniformity Grid Y cells

Specifies the vertical grid size (in pixels) used for defining regions of interest used in the non-uniformity calculation.

Analysis Results Table

A results table, formatted as a tab-delimited string, is returned with a row of data for each Bayer color plane. The maximum neighbor non-uniformity percentage is reported for each region in the image .

Bayer plane image

A drop list allows the selection of which Bayer plane will be displayed in the image window. The maximum neighbor non-uniformity percentage is overlaid on the image. 3D view is also available.

% Deviation Non-Uniformity Image

The percentage deviation non-uniformity data can also be displayed. 3D View is also available.