Color histogram

Figure 6.8 The input image is shown in (a). The graph in (b) shows the color histogram. A quantization of 10 was used for each color channel. Therefore, there are 1000 buckets in the histogram. Each bucket is represented by a cube where the size of the cube is proportional to the number of pixels of the same color in the original image.

Figure 13.1 The first row shows images of different fruits. A banana, an orange, and a peach are shown. The color histogram is shown in the second row. The last row shows the color histogram in three dimensions.

Figure 13.2 The first row shows the peach at different two different scales and in a rotated position. The color histograms are shown below each image. We see that the histogram is not quite invariant to changes in scale or orientation.

Comparison of color histograms does not necessarily have to be done in RGB space. Histograms can also be computed for a variety of color spaces or certain subspaces. Swain and Ballard computed the histograms using three opponent color axes red-green (RG), blue-yellow (BY), and black-white (BW). Let c = [cr, cg, cb]T be the color of the input pixel, then the transformed pixel coordinates c = [crg, Cby, q,w]r are given as... 

Fig. 4.6. Identical single-color histograms derived from quite different cell samples. The dual-parameter dot plots allow us to distinguish these two populations of cells.

M 21] [P 20] Mixing of an extended circular mixing chamber (diameter 2.5 mm, depth 40 pm) is complete after 55 s when using a 3 x 3 array of micro impellers at 600 rpm, as demonstrated by dye flow visualization and color histograms [32],... 

Fig. 9.42. Frequency of various coincident boundaries as a function of the E number (adapted from Watanabe (1989)). The data are for a fully annealed Fe-6.5 wt%Si alloy. The two distributions (different color histograms) correspond to samples with different textures.

Feature extraction process Evaluate each patch with texture and shape descriptor. As for texture we analyze feature of contrast, correlation, energy and homogeneity of each medical image. We also measure noise and blur for quality characteristics evaluation. For shape descriptor we use Hu moment invariant and Fourier transform analysis. Finally for color feature we used color histogram descriptor. 

Figure 17 Droplet probability distribution function for the total mass in each droplet size class, as measured with the PDI technique (blue dashed gray in the print version) and in the DSMC simulation (green light gray in the print version) for 250 bar inlet pressure. For reference, the red (solid dark gray in the print version) color histogram is the initial Rosin-Rammler droplet size distribution used in the inlet region. The left figure is from samples obtained at 80 mm from the nozzle exit and 38 mm from the central axis. The right figure is from samples obtained at 380 mm from the nozzle exit and 80 mm from the central axis.

Figure 20 Steady-state SALS and SANS patterns obtained as a function of the shear rates for PS 200/DOP 8.0 wt.% at 22°C. The SANS patterns obtained at different sample-to-detector distances L (35.7 and f 0 m) are presented. The color bars for scattering intensity are shown on logarithmic scale. The values attached on the color bars are the lower and upper limits (from fO to f 0 arbitrary units for SALS and from f0 to f 0 cm" for SANS) of the color histograms. The scale bars attached to the respective SANS patterns indicate a q scale of 5.63 x f 0 nm for 35.7 m and 2.00 x f 0 nm for f 0 m, whereas the one attached to the SALS pattern indicates a pscale of 3.63 x f 0 nm 7 Based on Salto, S. Hashimoto, T. Morfin, I. etal. Macromolecules2QQ2,35,445. ...

---