summaryrefslogtreecommitdiffstats
path: root/worksheets/a1_textrain.md
diff options
context:
space:
mode:
authorMatt Strapp <matt@mattstrapp.net>2021-09-20 18:15:14 -0500
committerMatt Strapp <matt@mattstrapp.net>2021-09-20 18:15:14 -0500
commit342403a02f8063903d0f38327430721d4d0ae331 (patch)
tree29d020a27bc16939c568dd4b29166566d1c0e658 /worksheets/a1_textrain.md
parentFix parenthesis (diff)
downloadcsci4611-submission-p1.0.tar
csci4611-submission-p1.0.tar.gz
csci4611-submission-p1.0.tar.bz2
csci4611-submission-p1.0.tar.lz
csci4611-submission-p1.0.tar.xz
csci4611-submission-p1.0.tar.zst
csci4611-submission-p1.0.zip
Diffstat (limited to 'worksheets/a1_textrain.md')
-rw-r--r--worksheets/a1_textrain.md164
1 files changed, 82 insertions, 82 deletions
diff --git a/worksheets/a1_textrain.md b/worksheets/a1_textrain.md
index 2f9c64f..834f2b1 100644
--- a/worksheets/a1_textrain.md
+++ b/worksheets/a1_textrain.md
@@ -1,82 +1,82 @@
-# Assignment 1 (Text Rain) Worksheet
-
-For the conceptual worksheets in this course, we'll provide a Markdown
-template from the shared-upstream repository. As described in the Canvas
-assignment handouts, you'll pull each Markdown template into your repository,
-directly edit your local copy with your answers, and commit and push your
-answers to GitHub in your `worksheets` folder of your repository. If you're
-unfamiliar with Markdown syntax, [check out this lovely guide provided by
-GitHub](https://guides.github.com/features/mastering-markdown/) before you get
-started.
-
-_Do not make a copy of the provided Markdown template and submit that instead._
-Our grading scripts will be looking for these modified files within your
-`worksheets` folder of your repository. Do not change the filenames, simply
-modify the contents.
-
-## Background
-
-By default, Processing uses the integer-based `0-255` convention to represent
-colors. For instance, bright, full-saturation red is represented as
-`color(255, 0, 0)`. Processing also supports grayscale colors; black is
-`color(0)` and white is `color(255)`. You may wish to look at the [color class
-documentation](https://processing.org/reference/color_.html) and/or the
-[tutorial explaining color in
-Processing](https://processing.org/tutorials/color/).
-
-Here are a couple of questions to get you thinking about how to work with
-pixel arrays and colors in this format. Note: These are very brief questions
-in this first worksheet, so this may not take you long at all. That's ok!
-
-
-## Q1: Indexing
-
-As mentioned in the assignment handout, accessing/setting pixel data via
-Processing's `get()` and `set()` routines is a bit easier to code, but it's
-much slower than directly accessing/changing a [PImage
-object's](https://processing.org/reference/PImage.html) `pixels[]` array.
-Processing stores a 2D image in this 1D array, so getting the proper pixel out
-requires a little additional math.
-
-In the code block below, write the equation for obtaining the index in the 1D
-array from a (row, column) in the 2D pixels array. Keep in mind you can use
-information from `inputImg` to help you.
-
-```java
-PImage inputImg = loadImage("test.jpg");
-
-// your code should work for any valid values for row and column, we've
-// randomly picked the values (2, 2) here as an exmaple.
-int row = 2;
-int column = 2;
-
-// write your answer in terms of the row and column defined above
-int index1D = row * inputImg.width + column;
-```
-
-
-## Q2: Thresholding
-
-The image processing technique known as *thresholding* will be useful while
-creating your Text Rain. During the thresholding operation, if a pixel's
-grayscale value is less than `threshold`, then it becomes black. If the
-value is greater than or equal to `threshold`, it becomes white. In the example below,
-assume the image has already been converted to grayscale. This means the
-red, green, and blue channels are all equal. So, you can get the grayscale
-value by accessing any one of the color channels red, green, or blue.
-
-In the code block below, write a Java code snippet for thresholding one pixel
-(`inputPixel`) to black or white.
-
-```java
-final int threshold = 128;
-
-// Returns: thresholded color (black or white)
-color thresholdPixel(color inputPixel) {
- if (blue(inputPixel) < threshold) {
- return color(0);
- } else {
- return color(255);
- }
-}
-```
+# Assignment 1 (Text Rain) Worksheet
+
+For the conceptual worksheets in this course, we'll provide a Markdown
+template from the shared-upstream repository. As described in the Canvas
+assignment handouts, you'll pull each Markdown template into your repository,
+directly edit your local copy with your answers, and commit and push your
+answers to GitHub in your `worksheets` folder of your repository. If you're
+unfamiliar with Markdown syntax, [check out this lovely guide provided by
+GitHub](https://guides.github.com/features/mastering-markdown/) before you get
+started.
+
+_Do not make a copy of the provided Markdown template and submit that instead._
+Our grading scripts will be looking for these modified files within your
+`worksheets` folder of your repository. Do not change the filenames, simply
+modify the contents.
+
+## Background
+
+By default, Processing uses the integer-based `0-255` convention to represent
+colors. For instance, bright, full-saturation red is represented as
+`color(255, 0, 0)`. Processing also supports grayscale colors; black is
+`color(0)` and white is `color(255)`. You may wish to look at the [color class
+documentation](https://processing.org/reference/color_.html) and/or the
+[tutorial explaining color in
+Processing](https://processing.org/tutorials/color/).
+
+Here are a couple of questions to get you thinking about how to work with
+pixel arrays and colors in this format. Note: These are very brief questions
+in this first worksheet, so this may not take you long at all. That's ok!
+
+
+## Q1: Indexing
+
+As mentioned in the assignment handout, accessing/setting pixel data via
+Processing's `get()` and `set()` routines is a bit easier to code, but it's
+much slower than directly accessing/changing a [PImage
+object's](https://processing.org/reference/PImage.html) `pixels[]` array.
+Processing stores a 2D image in this 1D array, so getting the proper pixel out
+requires a little additional math.
+
+In the code block below, write the equation for obtaining the index in the 1D
+array from a (row, column) in the 2D pixels array. Keep in mind you can use
+information from `inputImg` to help you.
+
+```java
+PImage inputImg = loadImage("test.jpg");
+
+// your code should work for any valid values for row and column, we've
+// randomly picked the values (2, 2) here as an exmaple.
+int row = 2;
+int column = 2;
+
+// write your answer in terms of the row and column defined above
+int index1D = row * inputImg.width + column;
+```
+
+
+## Q2: Thresholding
+
+The image processing technique known as *thresholding* will be useful while
+creating your Text Rain. During the thresholding operation, if a pixel's
+grayscale value is less than `threshold`, then it becomes black. If the
+value is greater than or equal to `threshold`, it becomes white. In the example below,
+assume the image has already been converted to grayscale. This means the
+red, green, and blue channels are all equal. So, you can get the grayscale
+value by accessing any one of the color channels red, green, or blue.
+
+In the code block below, write a Java code snippet for thresholding one pixel
+(`inputPixel`) to black or white.
+
+```java
+final int threshold = 128;
+
+// Returns: thresholded color (black or white)
+color thresholdPixel(color inputPixel) {
+ if (blue(inputPixel) < threshold) {
+ return color(0);
+ } else {
+ return color(255);
+ }
+}
+```