e. non-invertible.Figure 1.The TOMBO architectureTo overcome the above limitations, Tanida et al proposed a new image reconstruction approach called, pixel rearrange method [10], which could be integrated to enable the realization of a compact, low cost thin imaging system. In their approach, a cross-correlation based technique is used to arrange and align unit image pixels. To correct for the misalignment, a unit reference image is used. The relative shift values (��x and ��y in Fig. 2) of each unit image with respect to the reference image are determined by identifying the peak location of the cross-correlation function between the unit image and the reference one. Interpolation techniques were used to identify the cross-correlation peak [10].

The cross-correlation based pixel rearrange method is illustrated in Fig.

2.Figure 2.Cross correlation-based pixel rearrange methodIn the rearrangement process, it is assumed that the cross-correlation function is ideally symmetric around a single peak. In other words, there is a single shift between the considered unit image and the reference image, i.e., the spatial PSF function has only one parameter. In reality, however, there would be more than one parameter in a PSF (i.e., several cross-correlation peaks). This will limit the performance of the rearrange method when aligning unit images. Furthermore, the presence of several cross-correlation peaks introduces additional blur in the restored image. Inverse filtering is subsequently required.

This operation is not only computationally costly but also unstable if at least a single non-minimum phase component is present.

It also requires for the PSF to be known. Besides the spatial PSF, additive noise can also introduce false cross-correlation peaks, which further degrades significantly the performance of rearrange method [10].In the same paper, Tanida et al proposed a method to minimize the problems associated with: (i) TOMBO’s intrinsic PSFs (ii) imager internal noise, and (iii) shading introduced Drug_discovery by the separating walls (Fig. 1). To overcome these problems, Tanida et al introduced a de-shading pre-processing step, which uses a black picture and a white one for calibration.

We can analyze the de-shading process by noting that,B(x,y)=hint(x,y)Bi(x,y)+VB(x,y)(1)W(x,y)=hint(x,y)Wi(x,y)+VW(X,Y)(2)where, x and y define the pixel Entinostat location, hint(x, y) represents the intrinsic PSF of the TOMBO imager, Bi(x, y) and Wi(x, y) are the black and white pictures to be captured, B(x, y) and W(x, y) are the captured black and white images, and VB(x, y) and VW(x, y) are the additive internal noise for the black and white images respectively.By subtracting Eqn. (1) from Eqn.