4 (Part I) IME: Image Manipulation and EnhancementJava

Java Python Assignment 4 (Part I)

IME: Image Manipulation and Enhancement

1 Introduction

In this series of assignments, you will build an image processing application. You will create text-based and interactive GUI-based user interfaces for this program. Although you will implement a few simple and not-so-simple image processing effects, your system will have "good bones" through design that will streamline adding more image manipulations.

1.1 Images

Whatever the file format (jpg, png, bmp, etc.) an image is basically a sequence of pixels. Each pixel has a position in the image (row, column) and a color. For a greyscale image there is only one value per pixel. On a scale of 0-1, 0 indicates black and 1 indicates white. Values in between indicate shades of    grey. This value is traditionally represented using an integer. The number of bits used to store this value dictateshow many "levels of grey" are supported. For example, an 8-bit representation creates 256 distinct levels (including black and white).

1.1.1 Color Representation

For a color image,the pixel's color is represented by breaking it into individual components (usually 3) in  some way. The most common representation is the red-green-blue (RGB) model. In this model, a color is represented by three numbers (components): red, green, blue. Any color is a combination of these three base colors. On a scale of 0-1, black is represented as (0,0,0),white as (1,1,1) and bright, pure yellow is  represented as (1,1,0). There are several ways of measuring "brightness" or "intensity":

· Value: the maximum value of the three components for each pixel

· Intensity: the average of the three components for each pixel.

· Luma: the weighted sum 0.2126r+0.7152g+0.0722b

The three components can also be represented using integers. Each component is also called a

"channel". If 8 bits are used per channel, this creates the (traditionally-termed) 24-bit image! Some formats also support transparency (e.g. the PNG format): this is done by adding a fourth component.

To see how various colors can be made using these base colors: open Intellij -> Type "Ctrl+Shift+A" on Windows or "Command+Shift+A" on Mac and type "show Color Picker". When selected, you can pick   different colors and observe their red, green and blue values to get an intuition for how this works).

1.2 Image Processing

The ubiquity of images is matched by the plethora of image manipulation programs and services. All phones with a camera also include apps that allow us to brighten, darken, blur or crop photos. Instagram has''filters" that convert a picture into something more interesting. Programs on traditional computers range from very simple photo manipulators to sophisticated Photoshop-like applications. The technical field of image processing itself is deeply rooted in math and signal processing theory. But the basics of image representation and manipulation are quite simple to understand. All image processing applications boil down to manipulating individual pixel colors. Some operations require computing properties or statistics on an image. For example, to increase the contrast in an image, one has to ensure that the intensities of pixels span a wider range (and hence can be thought of as an operation on the histogram of an image).

1.3 Sample Image Application

If you have not worked with image processing applications before, we recommend

Gimp(http://www.gimp.org). Gimp is a free, open-source powerful image processing application. If you have access to Photoshop then that will be enough as well. We encourage you to find and tryout the  image processing operations that we will introduce in these assignments using these applications, to understand them better.

2 A sampling of image processing operations

2.1 Example Image Processing: Visualizing components

One of the easiest ways to analyze a color image is to separately inspect the components (channels) of  its pixels. This provides insight into the overall color balance of the image, or which components seem to store more information than others. This information can be used to compress the image better (by discarding data from or using fewer bits for its less-important channels) or make image processing faster (by only working on some channels).

The easiest way to visualize the channel of an image is to create an image, where the red, green and blue values for a pixel are all set to the value of a particular channel. This creates a greyscale visualization of a channel. For example, if a pixel in the