Software Settings

Image Parameters

Common Terms

Name	Value Range	Impact Factors
Image Quality	1 ~ 100	Lower values mean lower quality
Alpha Quality	0 ~ 100	For images containing Alpha channels (such as PNG or certain PSD formats), adjusting Alpha plane fidelity means controlling the degree of Alpha channel information loss during compression. Higher fidelity settings retain more details, but file size will be relatively larger; lower fidelity settings sacrifice some details for smaller file size. Lower values mean lower quality
Bit Depth	1 ~ 16	Bit Depth, in image processing, refers to the number of bits used per pixel, which determines the number of colors an image can represent. Higher bit depth means richer colors and smoother color transitions. For example, 8-bit depth can represent 256 colors, while 16-bit depth can represent 65536 colors. Lower values mean smaller file size, but lower color quality. It’s recommended to adjust moderately or keep unchanged
Subsampling Mode (Chroma Subsampling)	ON/OFF/AUTO	In image and video encoding, chroma subsampling is a technique to reduce data volume by lowering the resolution of chroma (color) information. This is because the human eye is more sensitive to brightness changes than chroma changes.
Preserve Metadata	Preserve/Don’t Preserve	Metadata is data about data, which can provide detailed information about files, including but not limited to: - Creation date and time - Modification date and time - Device information (such as camera model, lens information) - Image parameters (such as resolution, color space, exposure settings) - Geographic location information (GPS coordinates) - Copyright information - Descriptive tags or keywords In some cases, to protect privacy, reduce file size, or simplify data management, users may need to choose to preserve part or all of the metadata, or completely delete certain types of metadata. This option usually appears in media processing software or device settings menus.
Lossless Compression	-	Lossless compression is a data compression method where the original data can be completely restored, meaning the data before and after compression are completely identical with no information loss. This compression method is commonly used in scenarios that require maintaining data integrity, such as text files, program code, certain types of images (such as professional photography or scientific imaging), and audio files. When lossless compression is enabled, the software or device uses algorithms to find repeated patterns or redundant information in the data and stores this information in a more efficient way, thereby reducing file size without sacrificing data quality or integrity. Once needed, this data can be decompressed back to its original state without any changes. Lossless compression preserves original data. For smaller file size, do not check this option.
Progressive Image	Boolean	In traditional baseline JPEG, images are decoded and displayed line by line from top to bottom. This means that before the image is fully loaded, you can only see the top part, and the bottom part is still blank. In progressive JPEG, the image is first displayed at very low resolution, i.e., a very blurry version. As more data is loaded, the image’s clarity gradually increases until it is finally fully clear. The advantage of this method is that even with slow network connections, users can quickly preview the general content of the image, improving user experience.
Dithering	0~1

WebP

Name	Value Range	Impact Factors
Smart Chroma Subsampling	Boolean	Intelligently adjusts chroma subsampling
Smart Minimum Size	Boolean	Optimize to achieve minimum file size
Mixed Encoding	Boolean	In image or video encoding, different regions or frames may have different characteristics, such as static backgrounds with dynamic foregrounds, texture-rich areas with smooth gradients, etc. Mixed encoding strategies can intelligently select the most suitable encoding method based on content characteristics, such as using lossless compression for static parts and lossy compression for dynamic parts, or combining multiple lossy compression parameter settings to achieve the best overall compression ratio

JPEG

Name	Value Range	Impact Factors
Use Predefined Quantization Table with Given Index	0 ~ 8	Different quantization tables can produce different degrees of compression and corresponding quality loss. Predefined quantization tables are designed in advance, optimized quantization matrix sets based on specific application scenarios and quality requirements. Using predefined quantization tables allows selecting the appropriate quantization table through a specified index to achieve the expected compression effect and image quality balance. 0 - Quantization table from JPEG Appendix K (default used by vips and libjpeg) 1 - Flat quantization table 2 - Quantization table optimized for MSSIM metric on Kodak image set 3 - Quantization table designed by N. Robidoux for ImageMagick (currently default for mozjpeg) 4 - Quantization table optimized for PSNR-HVS-M metric on Kodak image set 5 - Quantization table from the paper “Correlation Between Human Visual Perception and JPEG-DCT Compression” (1992) 6 - Quantization table from the paper “DCTune: Perceptual Optimization of Compressed Dental X-rays” (1997) 7 - Quantization table from the paper “Application of a Visual Detection Model to DCT Coefficient Quantization” (1993) 8 - Quantization table from the paper “An Improved Visual Detection Model for DCT Coefficient Quantization” (1993) Each quantization table has its own characteristics and is optimized for different image quality and compression efficiency.
Add Restart Markers After Specified Number of MCUs	0 - 2147483647	In image compression standards such as JPEG, MCU is the smallest encoding and decoding unit, usually containing one luminance block and two chrominance blocks, with its size depending on the chroma subsampling mode used. Restart markers are special markers in the JPEG format that are inserted into the image data stream as checkpoints during image encoding. These markers can help the decoder resynchronize from the nearest restart marker when encountering errors (such as data corruption or transmission interruption), rather than decoding the entire image from the beginning. This improves decoding robustness and fault tolerance, especially in scenarios with unstable network transmission or storage media. By specifying how many MCUs to add a restart marker after, a balance can be found between image quality and decoding robustness. Increasing the number of restart markers can improve decoding robustness, but may slightly increase file size, as each marker occupies a certain number of bytes.
Optimal Huffman Coding Table	Boolean	Based on the actual statistical characteristics of image data, calculate a Huffman coding table that maximizes compression ratio. This process is usually performed during the encoding phase by analyzing image data, counting the frequency of each DCT coefficient, then constructing a Huffman tree and generating a coding table. During decoding, the decoder uses the same coding table to restore the original data, thus completing lossless decompression.
Apply Trellis Quantization to Each 8x8 Block	Boolean	To improve encoding efficiency and image quality, apply an optimized quantization strategy—trellis quantization—to each 8x8 pixel image block. This technique improves overall encoding efficiency by making fine-grained decisions at each quantization step to achieve optimal rate-distortion performance.
Apply Overshoot to Samples with Extreme Values	Boolean	When processing signals or images, intentionally increase overshoot effects for points with particularly large or small values to enhance local features, but care must be taken to control the degree of overshoot to avoid image quality degradation.
Split DCT Coefficient Spectrum into Separate Scans	Boolean	In JPEG image compression, split the spectrum of Discrete Cosine Transform (DCT) coefficients into multiple independent scan processes. In JPEG encoding, images are first divided into 8x8 pixel blocks, and after DCT transformation of each block, a series of DCT coefficients are obtained. These coefficients contain frequency information of the image, which is particularly important in progressive JPEG encoding, allowing images to gradually become clear during download, first displaying low-frequency information (basic outline), then gradually adding more details. This allows users to preview the general content of the image faster, improving user experience, especially with slow network speeds.

PNG

Name	Value Range	Impact Factors
Compression Factor	0~9	Lower values mean lower quality
libspng Row Filter Flags	none, sub, up, avg, paeth, all	In PNG image compression, filter preprocessing is an important step to improve compression efficiency. The above describes different filter types supported by the PNG standard: NONE: Do not use any filter, i.e., no preprocessing, directly compress raw data. SUB: Use the difference from the left neighboring pixel for prediction. Each pixel value subtracts the value of the left pixel to reduce data redundancy. UP: Use the difference from the pixel above for prediction. Each pixel value subtracts the value of the pixel directly above, also to reduce correlation between data. AVG: Use the average of the left neighbor and the pixel above for prediction. Each pixel value subtracts the average of the two pixels in the upper left corner, combining information from both horizontal and vertical directions. PAETH: Use the Paeth predictor to automatically select the best neighboring pixel for prediction. This is a more complex prediction algorithm that selects the most appropriate prediction value based on the specific situation of surrounding pixels to achieve better compression. ALL: Adaptive filter selection. During encoding, PNG will try all five filters and select the filter type with the best compression effect for each row. This strategy usually provides the highest compression ratio, but the computational cost is relatively higher. The role of these filters is to reduce correlation between adjacent pixels before encoding, so that subsequent entropy encoding (such as DEFLATE compression) can compress data more effectively and improve compression ratio.
Use 8bpp Palette Quantization	false	Quantize the image to use 8 bits per pixel (bpp) palette color mode. This means the entire image’s colors will be limited to a palette containing at most 256 colors. This method is very suitable for images with a limited number of colors that do not require high color fidelity, such as simple graphics, icons, or certain types of cartoons. Using 8bpp palette in PNG can achieve efficient lossless compression while maintaining good visual quality, especially on images with limited colors.

GIF

Name	Value Range	Impact Factors
Maximum Inter-frame Error	0~32	”Maximum inter-frame transparency error” refers to the maximum allowed difference in transparency changes from one frame to the next in multi-frame animations. This is important in the compression process because GIF animations are usually composed of a series of individual frames, each of which may have its own transparent areas. To reduce file size, similar parts between adjacent frames are efficiently encoded, storing only those pixels that have changed. In practical applications, creators need to adjust this parameter according to specific needs to achieve the best compression and quality balance.
Maximum Error Between Palettes	0~256	When considering palette reuse, animation compression algorithms compare the color requirements of the current frame with the palette of the previous frame (or other available frames) to determine the degree of match. This matching degree can be quantified by calculating the “error” between the two palettes, where the error usually refers to the sum of color differences. If the difference between two palettes is small enough that they are visually almost indistinguishable, the previous palette can be reused, thus saving space.
Reuse Palette	true	Helps reduce file size by avoiding duplicate storage of similar palettes. When processing consecutive two or more frames, if they share many of the same colors, reusing the previous frame’s palette can significantly save space.

TIFF/JP2K

Name	Value Range	Impact Factors
Tile Width (in pixels)	1- 32768	In TIFF format, tile width refers to the horizontal dimension of tiles that make up the image, i.e., the number of pixels in a tile in the horizontal direction. Tiles are rectangular regions into which the image is divided for block storage and reading, facilitating the processing of large images.
Tile Height (in pixels)	1-32768	Tile height refers to the number of pixels contained in a tile in the vertical direction. Together with tile width, these two parameters define the size of a single tile.
Horizontal Resolution (pixels/mm)	0.001 ~ 1e+06	Horizontal resolution represents the number of pixels contained in the image per millimeter of horizontal distance. This is a measure of image detail density. Higher horizontal resolution means more pixels in the same physical space, with richer image details.
Vertical Resolution (pixels/mm)	0.001 ~ 1e+06	Vertical resolution similarly measures the number of pixels per millimeter of vertical distance. It reflects the degree of detail in the vertical direction of the image, and together with horizontal resolution, determines the overall clarity and detail expression of the image.
Write Pyramid TIFF	Boolean	This is a multi-level format for storing images, containing multiple copies of the image at different resolutions, from full resolution to lower resolutions. This format is suitable for application scenarios that require quick preview or scaling, as it allows quick access to different levels of detail.
Write Tiled TIFF	Boolean	Tiled TIFF divides the image into multiple tiles or blocks, each of which can be independently compressed and stored. This method improves the read and write efficiency of large images, especially for random access and network transmission.
Use 0 to Represent White in 1-bit Images	Boolean	In 1-bit images (binary images), each pixel is represented by only one bit, usually 0 represents one color and 1 represents another color. Traditionally, 0 is used to represent black, while 1 represents white. However, in some cases, as mentioned, 0 can be configured to represent white, which is a configurable convention in image processing and encoding. This configuration affects how image data is decoded to correctly display the image.

HEIF

Name	Value Range	Impact Factors
Compression Format	hevc, avc, jpeg, av1	In HEIF (High Efficiency Image File Format) compression, the above lists several different compression algorithm options, which are: HEVC: Use HEVC (High Efficiency Video Coding, also known as H.265) standard for compression. HEVC is an efficient image and video compression format that provides higher compression efficiency compared to previous standards. x265 is an open-source library for HEVC encoding. AVC: Use AVC (Advanced Video Coding, also known as H.264) standard for compression. AVC is a widely used video compression standard that balances compression efficiency with computational resource requirements. x264 is a popular AVC encoder. JPEG: Use JPEG standard for compression. JPEG is a lossy compression format widely used for static images, known for maintaining good image quality at smaller file sizes. The standard JPEG encoder is used here. AV1: Use AV1 (AOMedia Video 1) standard for compression. AV1 is a next-generation video codec developed by the Alliance for Open Media, designed to provide higher compression efficiency than HEVC while maintaining open-source licensing. aom is a library led by Google for AV1 encoding.
Encoder	auto, aom, rav1e, svt, x265	In HEIF (High Efficiency Image File Format) compression, the above lists different encoder options used to compress image data into HEIF-supported formats. Specific explanations are as follows: AUTO: Indicates automatic encoder selection. In this mode, the system or software will automatically select the most suitable encoder based on specific circumstances and available resources. AOM: Refers to using the encoder developed by AOMedia (AOM), mainly for AV1 video encoding. AOMedia is an alliance led by Google, and its AV1 encoder is designed to provide efficient compression. RAV1E: RAV1E is an open-source AV1 encoder known for high performance and high-quality video compression. It is a community-driven project focused on providing efficient AV1 encoding solutions. SVT: SVT-AV1 is another AV1 encoder jointly developed by Socionext Inc. and Amazon Lab126. It is designed to provide efficient video compression suitable for various application scenarios. X265: x265 is an efficient HEVC (H.265) video encoder widely used for compressing HD and UHD videos. It is known for providing excellent compression ratio and image quality. Choosing different encoders in HEIF compression affects the final file size, encoding speed, decoding compatibility, and image quality. Users can select the most suitable encoder based on specific needs and device capabilities.

AVIF

Name	Value Range	Impact Factors
-	-	-

Golang Quick Reference macOS CI/CD Apple Signing Cheatsheet