Computer vision basics quiz

Question 1

What is a digital image at its most basic level, and what is a pixel?

Accepted Answer

A 2D grid of pixels, each storing a numeric intensity/color value

Answer

A continuous analog signal with no discrete units

Answer

A list of vector shapes defined by equations

Answer

A single number describing the whole scene

Question 2

In an 8-bit grayscale image, what range of values can each pixel take and what do the extremes mean?

Accepted Answer

0 to 255, where 0 is black and 255 is white

Answer

0 to 100, where 0 is white

Answer

-128 to 127, where 0 is gray

Answer

1 to 256, where 1 is white

Question 3

What are the three channels in an RGB image, and how is a color formed from them?

Accepted Answer

Red, Green, Blue combined additively

Answer

Cyan, Magenta, Yellow combined subtractively

Answer

Hue, Saturation, Value combined multiplicatively

Answer

Black, White, Gray combined by averaging

Question 4

Why are images often converted to grayscale before certain computer vision tasks?

Accepted Answer

To reduce data and focus on intensity/structure when color is not essential

Answer

Because color images cannot be stored on computers

Answer

Because grayscale always improves accuracy for every task

Answer

To increase the number of channels for more detail

Question 5

What is the shape of a typical color image array in libraries like NumPy/OpenCV, and what does each dimension mean?

Accepted Answer

(height, width, channels), e.g. (480, 640, 3)

Answer

(channels, width, height) always in RGB order

Answer

(width, height) with no channel dimension

Answer

(channels, height) only

Question 6

What is an image histogram and what information does it convey?

Accepted Answer

A count of pixels at each intensity level, showing brightness distribution

Answer

A map of object locations in the image

Answer

The list of edges detected in the image

Answer

The compressed file size of the image

Question 7

What does histogram equalization do to an image?

Accepted Answer

Spreads intensities to increase contrast

Answer

Blurs the image to remove noise

Answer

Detects edges by finding gradients

Answer

Converts color to grayscale

Question 8

What is image convolution, and what is a kernel (filter)?

Accepted Answer

Sliding a kernel over the image and computing weighted sums of neighbors

Answer

Reordering pixels randomly across the image

Answer

Compressing the image into a smaller file

Answer

Converting pixels to vector graphics

Question 9

What is the effect of a Gaussian blur filter and why is it commonly used before edge detection?

Accepted Answer

It smooths and reduces noise using a weighted neighbor average

Answer

It sharpens edges by amplifying differences

Answer

It increases image resolution

Answer

It converts the image to binary

Question 10

What is an edge in an image, and what local property identifies it?

Accepted Answer

A location of rapid intensity change (high gradient)

Answer

A region of constant intensity

Answer

The corner where two image files meet

Answer

The brightest pixel in the image

Question 11

What is the role of the Sobel operator in computer vision?

Accepted Answer

Estimates intensity gradients to detect edges and their direction

Answer

Compresses images using gradients

Answer

Equalizes the histogram

Answer

Resizes the image while keeping aspect ratio

Question 12

Outline the main steps of the Canny edge detector.

Accepted Answer

Smooth, compute gradients, non-maximum suppression, hysteresis thresholding

Answer

Threshold, rotate, crop, save

Answer

Resize, blur, sharpen, invert

Answer

Convert to RGB, equalize, blur, output

Question 13

What is thresholding in image processing, and what is binary thresholding?

Accepted Answer

Comparing each pixel to a cutoff to produce a binary image

Answer

Mixing color channels together

Answer

Rotating the image by a fixed angle

Answer

Counting the total number of pixels

Question 14

What problem does Otsu's method solve in thresholding?

Accepted Answer

It automatically selects an optimal threshold from the histogram

Answer

It increases image resolution

Answer

It detects faces in an image

Answer

It compresses the image losslessly

Question 15

What is image segmentation in computer vision?

Accepted Answer

Partitioning an image into meaningful regions or per-pixel labels

Answer

Assigning a single label to the whole image

Answer

Compressing the image into segments for storage

Answer

Splitting a video into individual frames

Question 16

Distinguish between image classification, object detection, and segmentation.

Accepted Answer

Classification labels the whole image, detection adds bounding boxes, segmentation labels pixels

Answer

They are three names for the exact same task

Answer

Detection labels pixels while segmentation draws boxes

Answer

Classification always uses video, the others use images

Question 17

What is a bounding box in object detection, and how is it typically represented?

Accepted Answer

A rectangle enclosing an object, given by coordinates like (x, y, w, h)

Answer

A circle covering the whole image

Answer

A single pixel marking the object center only

Answer

A color histogram of the object

Question 18

What does Intersection over Union (IoU) measure in object detection?

Accepted Answer

Area of overlap divided by area of union of two boxes

Answer

The brightness difference between two boxes

Answer

The number of pixels inside a box

Answer

The confidence score of a detection

Question 19

Why are Convolutional Neural Networks (CNNs) well suited to images compared to fully connected networks?

Accepted Answer

They use weight-sharing filters that reduce parameters and capture local patterns

Answer

They ignore spatial structure entirely

Answer

They require no training data

Answer

They only work on grayscale images

Question 20

What is the purpose of a pooling layer (e.g., max pooling) in a CNN?

Accepted Answer

To downsample feature maps, reducing size while keeping key information

Answer

To add more pixels and increase resolution

Answer

To convert features into RGB colors

Answer

To randomly shuffle the feature values

Question 21

In a CNN, what kinds of features do early layers versus deep layers typically learn?

Accepted Answer

Early layers learn edges/simple patterns, deep layers learn complex parts/objects

Answer

Early layers learn whole objects, deep layers learn edges

Answer

All layers learn the same features

Answer

Only the final layer learns any features

Question 22

What is a feature map (activation map) in a CNN?

Accepted Answer

The output of a filter showing where its pattern appears in the input

Answer

A map of the original raw pixel colors

Answer

The list of class labels predicted

Answer

A compressed copy of the input file

Question 23

What is padding in convolution, and why is 'same' padding used?

Accepted Answer

Adding border pixels so output size can match input size

Answer

Removing border pixels to shrink the image

Answer

Doubling each pixel's value

Answer

Converting the image to grayscale

Question 24

What is stride in a convolution or pooling operation, and how does it affect the output size?

Accepted Answer

The step size the filter moves; larger stride gives smaller output

Answer

The number of color channels

Answer

The brightness of each pixel

Answer

The number of training epochs

Question 25

Calculate the output size of a convolution: given a 32x32 input, a 5x5 filter, stride 1, and no padding, what is the output spatial size?

Accepted Answer

28x28

Answer

32x32

Answer

27x27

Answer

30x30

Question 26

What is data augmentation in computer vision and why is it useful?

Accepted Answer

Applying transformations to images to expand data and reduce overfitting

Answer

Deleting hard examples from the dataset

Answer

Increasing the learning rate during training

Answer

Compressing images to save disk space

Question 27

What is transfer learning, and why is it widely used in computer vision?

Accepted Answer

Reusing a pretrained model and adapting it to a new task

Answer

Training a model from scratch with random weights

Answer

Transferring images between computers

Answer

Converting a model to run on a phone

Question 28

What is the difference between RGB and HSV color spaces, and when is HSV useful?

Accepted Answer

HSV separates hue from brightness, easing color-based detection under varying light

Answer

HSV and RGB are identical representations

Answer

RGB always gives better color segmentation than HSV

Answer

HSV stores depth instead of color

Question 29

What is image resizing/interpolation, and why does downsampling risk losing information?

Accepted Answer

Downsampling discards pixels, permanently losing fine detail

Answer

Resizing never loses any information

Answer

Interpolation adds true new detail to the image

Answer

Resizing only changes file format, not pixels

Question 30

What is the difference between salt-and-pepper noise and Gaussian noise, and which filter handles salt-and-pepper best?

Accepted Answer

A median filter handles salt-and-pepper noise best by removing outliers

Answer

A Gaussian blur removes salt-and-pepper noise best

Answer

Salt-and-pepper noise cannot be removed at all

Answer

Sharpening removes salt-and-pepper noise best

Test yourself