Normalizing the background and removing the trend in one-dimensional DNA fingerprint images

Automatic segmentation of DNA bands in one dimensional gel images produced by hybridizing techniques

Automatic segmentation of DNA bands in one dimensional DNA fingerprint gel images produced by hybridizing techniques could be challenging due to the noise and large intensity variation among the bands within each lane in the image. A new algorithm for automatic segmentation of DNA bands based on the variance (var), mean-variance-ratio (MVR) and equivalent width is presented. A number of pre-processing steps should be applied on the image prior to the automatic segmentation. First a homomorphic filter is applied to enhance the image followed by an edge preserving noise filtering algorithm to remove the noise in the image. Then a background normalization operation is applied to the resulting image. After these preprocessing operations, each lane in the image is detected and presented by a one dimensional intensity profile. The new algorithm will be applied on the valleys in the resulting intensity profile for each lane to characterize the corresponding valley as noise or a DNA band. We have applied this algorithm on several DNA fingerprint gel images, and all the bands were successfully segmented, except the very weak (faint) bands which are very close to their background. These should be analyzed manually.

Quantitative evaluation of thin-layer chromatography with image background estimation based on charge-coupled device imaging

The main limitation of image processing based on charge-coupled device (CCD) imaging in quantitative evaluation of thin-layer chromatography (TLC) is its nonuniform illumination. This paper presents a novel method to solve the problem. First, the background is estimated by employing the cubic spline surface construction. Then the estimated background is subtracted from the original image to make the image segmentation; Finally, the gray level integration for each object is calculated, which is proportional to the value of samples. Experiment shows that the linear correlation coefficient between gray integration and concentration of samples is 0.9978.