Image Processing Lecture Review

Vocabulary-style flashcards covering fundamental concepts of image processing, TV standards, human vision, color models, transformations, and compression based on the provided practice questions.

Image

Information about objects or illuminated scenes that humans observe and perceive using their eyes and visual nervous system.

Still Image representation

Represented by a luminance function of coordinate variables in the image plane $I(x,y)$.

Sequence of Images representation

Represented by a luminance function of coordinate variables and time $I(x,y,t)$, where the temporal relationship between frames represents motion (Video).

Bit depth - Binary Image

Typically 1 bit/pixel.

Bit depth - Gray Image

Typically 8 bits/pixel.

Bit depth - Color Image

Typically ranges from 16 to 24 bits/pixel.

NTSC Standard

A television standard using 525 scan lines and a rate of 30 frames/second.

PAL Standard

A television standard using 625 scan lines and a rate of 25 frames/second.

SECAM Standard

A television standard featuring 625 scan lines and 25 frames/second.

Traditional Cinema Standard

Uses a frame rate of 24 frames/second.

HDTV Standard (16:9)

Described as having 720 scan lines and a rate of 60 frames/second.

SVGA Standard

Described with a resolution of $1024 \times 768$ pixels and a rate of 60 frames/second (or $1024 \times 720$ at 72 frames/second in some contexts).

Image Digitization

A process consisting of two basic steps: Sampling (Lấy mẫu) and Quantization (Lượng tử hóa).

Visible Light Spectrum

Electromagnetic waves ranging from approximately 380 nm to 780 nm (or 400 nm to 700 nm).

Rods (Tế bào que)

75-150 million cells that are very sensitive to light, responsible for night vision and perceiving luminance (brightness).

Cones (Tế bào nón)

6-7 million cells concentrated mainly in the yellow spot (fovea centralis), responsible for color sensitivity.

Luminance (Độ rọi)

A measure of light energy received from a light source, depending on distance and wavelength.

Radiance (Độ chói)

A physical measure proportional to the total energy emitted by a light source surface.

Brightness (Độ sáng)

A subjective attribute characterizing a human's ability to perceive light intensity.

Hue (Sắc độ)

A color attribute related to the primary wavelength in a mixture of light waves, representing the dominant perceived color.

Saturation (Độ bão hòa)

Represents relative purity, indicating the amount of white light mixed with the hue.

Chrominance

The collective term for the color characteristics Hue and Saturation.

CMYK Color Model

A subtractive color model used in printing where 'K' represents Black.

YCbCr Color Model

A color model used in compression; it separates the Luma (Y) from Chrominance (Cb for blue-difference, Cr for red-difference).

4-Neighborhood ($N_4(p)$)

The set of points at coordinates $(x+1,y)$, $(x-1,y)$, $(x,y+1)$, and $(x,y-1)$, relative to point $p(x,y)$.

Diagonal Neighborhood ($N_D(p)$)

The four diagonal points surrounding point $p(x,y)$: $(x+1,y+1)$, $(x-1,y-1)$, $(x+1,y-1)$, and $(x-1,y+1)$.

m-adjacency

Introduced to eliminate multiple paths or ambiguity in connectivity found in 8-adjacency.

Nyquist Sampling Theorem

States for a band-limited function, the maximum sampling intervals must satisfy $\text{Delta x} \times \frac{1}{2u_0}$ and $\text{Delta y} \times \frac{1}{2v_0}$.

Aliasing (Hiện tượng răng cưa)

Occurs when the sampling frequency does not satisfy the Nyquist rate (under-sampling).

Nearest Neighbor Interpolation

A fast interpolation method that assigns the value of the nearest pixel, though it may cause aliasing.

Bilinear Interpolation

Calculates new pixel values based on the 4 nearest neighbor pixels using a weighted average.

Bicubic Interpolation

Calculates new pixel values based on 16 nearest neighbor pixels, offering higher quality than Bilinear or Nearest Neighbor.

Histogram

A discrete function $h(r_k) = n_k$ representing the number of pixels $n_k$ with gray level $r_k$.

Histogram Equalization

A process to achieve a new image with an ideally uniform histogram distribution.

Salt & Pepper noise

Impulse noise appearing as black and white dots; negative impulses represent 'pepper' and positive impulses represent 'salt'.

Median Filter

An order-statistic filter that is highly effective at removing Salt & Pepper noise.

KL Transform (Karhunen-Loeve)

A transform aimed at reducing redundancy by minimizing correlation between neighboring pixels; its covariance matrix is a diagonal matrix.

Canny Edge Detection

A multi-stage edge detection method involving Gaussian smoothing, gradient calculation, non-maximum suppression, and hysteresis thresholding.

Hough Transform

A technique used to connect edge points into global shapes like lines, often using parameters such as $r$ and $\theta$.

Morphological Erosion ($A \text{ co } B$)

Defined as $A \text{ co } B = \{z | (B)_z \text{ is a subset of } A\}$, shrinking objects.

Morphological Dilation ($A \text{ dãn } B$)

Defined as $A \text{ dãn } B = \{z | (B)_z \text{ intersects } A \text{ and is not empty}\}$, expanding objects.

Opening (Phép mở)

A morphological operation consisting of Erosion followed by Dilation ($(A \text{ co } B) \text{ dãn } B$), used to remove small noise objects.

Closing (Phép đóng)

A morphological operation consisting of Dilation followed by Erosion ($(A \text{ dãn } B) \text{ co } B$), used to bridge small gaps and fill holes.

Huffman Coding

A lossless coding method using variable-length codewords where frequently occurring symbols get shorter codes, characterized by the prefix property.

JPEG Compression

A lossy compression standard utilizing 2D-DCT on 8x8 pixel blocks; at very high compression ratios, it may produce 'blocking artifacts'.