report #2


images partaken





background info

A binary image is an image of gray scale intensities of 0( black ) or 255( white ).

A threshold value is the pivotal point between 0 and 255 on the image.

A threshold value can best be determined using a histogram.

A histogram contains the intensity or the usage of each pixel value. These pixel values fall in the range of 0 to 255.

Any pixel value above this threshold value will be white and anything else below will become black.

 binary images with variable thresholds













discussion 1:

Discuss how good the binary image, derived from using threshold, is from the image histogram. Compare performance of various thresholds.

As one can see, at a high threshold value the image becomes dark; too dark for any effective deciphering of information. As the threshold increases to mid range values or to values where the intensity of the images pixels is most present, a more detail binary image is obtained. Thus more minute details can be seen.

 As the threshold goes toward the lower end of the spectrum of gray( closer to zero ),the image becomes more washed out or lighter since most pixel values will become 255 or white.

The best performance of a binary image comes when the threshold value gets chosen among a cluster of intensities. These intensities are found within the images histogram. Lennas better binary image came at a threshold value of 150, about mid range. Among the four possible choices above, the storms binary image best displayed itself using a threshold value of around 100.


global equalization

can you tell?





local equalization



local threshold


discussion 2:

Discuss histogram equalization results and improvement gained from local histogram equalization.

If an image has a cluster of pixel intensities huddled in a certain area, then the equalization techniques can spread out the cluster making full use of the gray scale spectrum.

For example, if an image has a darker band of intensities then the equalization can spread it out making the image a little lighter. The same holds true for the other end. That is, if an image is more flushed or brighter then equalization can add some more contrast to the original image.

If an image contains a "spot" possibly from a shadow or sun glare then one can use the technique of local equalization on the specific area; thus, keeping most of the original information of the image intact. For the lax image the global equalization would have been a better choice since the whole image is somewhat of a uniform mid-range gray.