processor

Documentation

Index

• func ExtractPalette(r io.Reader) ([]string, error)
• func GenerateAnimatedThumbnail(videoURL string, duration string, width int, height int, format string) (*bytes.Buffer, error)
• func GenerateStoryboard(videoURL string, interval string, cols, rows int, width int) (*bytes.Buffer, error)
• func GenerateThumbnail(videoURL string, timestamp string) (*bytes.Buffer, error)
• func LoadCascade(path string) error
• func Process(ctx context.Context, r io.Reader, opts ImageOptions, wmImg image.Image, ...) (*bytes.Buffer, error)
• func SmartCrop(img *vips.ImageRef, width, height int, detector ObjectDetector) error
• type AiDetector
  • func (d *AiDetector) Detect(img *vips.ImageRef) (*image.Rectangle, error)
• type EntropyDetector
  • func (d *EntropyDetector) Detect(img *vips.ImageRef) (*image.Rectangle, error)
• type ImageOptions
• type ObjectDetector

Functions

func ExtractPalette

func ExtractPalette(r io.Reader) ([]string, error)

ExtractPalette extracts dominant colors from the image.

func GenerateAnimatedThumbnail

func GenerateAnimatedThumbnail(videoURL string, duration string, width int, height int, format string) (*bytes.Buffer, error)

GenerateAnimatedThumbnail generates a 3-second animated thumbnail for a video file using ffmpeg. It extracts 3 seconds from the beginning (or timestamp).

func GenerateStoryboard

func GenerateStoryboard(videoURL string, interval string, cols, rows int, width int) (*bytes.Buffer, error)

GenerateStoryboard generates a storyboard image (grid of frames) for the video. interval: timestamp interval between frames (default "1") cols, rows: grid dimensions

func GenerateThumbnail

func GenerateThumbnail(videoURL string, timestamp string) (*bytes.Buffer, error)

GenerateThumbnail generates a thumbnail for a video file using ffmpeg. It returns a buffer containing the image data (JPEG).

func LoadCascade

func LoadCascade(path string) error

LoadCascade loads the pigo cascade file from the given path.

func Process

func Process(ctx context.Context, r io.Reader, opts ImageOptions, wmImg image.Image, wmOpacity float64, originalKey string) (*bytes.Buffer, error)

Process decodes, transforms, watermarks, and encodes the image. Note: We cannot easily pass context here without changing the signature, but typically Process is called from HandleRequest which has a context. Ideally Process should take context. For now we use Background if we can't change signature, BUT looking at where Process is called, it might be inside HandleRequest which likely has context.

func SmartCrop

func SmartCrop(img *vips.ImageRef, width, height int, detector ObjectDetector) error

SmartCrop applies the smart crop logic.

Types

type AiDetector

type AiDetector struct {
	ModelPath string
}

AiDetector uses ONNX Runtime to detect objects. It requires an ONNX model file path (e.g. YOLOv8n) and the ONNX Runtime shared library.

func (*AiDetector) Detect

func (d *AiDetector) Detect(img *vips.ImageRef) (*image.Rectangle, error)

type EntropyDetector

type EntropyDetector struct{}

EntropyDetector uses Shannon entropy to find the most "interesting" part of the image.

func (*EntropyDetector) Detect

func (d *EntropyDetector) Detect(img *vips.ImageRef) (*image.Rectangle, error)

type ImageOptions

type ImageOptions struct {
	Width            int
	Height           int
	Fit              string // cover, contain, fill, inside
	Format           string // jpeg, png, webp, jxl
	Quality          int
	Focus            string // smart, face
	Text             string
	TextColor        string
	TextSize         float64
	TextOpacity      float64
	Font             string
	Effect           string
	Brightness       float64
	Contrast         float64
	Blurhash         bool
	SmartCompression bool
	Animated         bool
	Page             int
}

type ObjectDetector

type ObjectDetector interface {
	// Detect returns a crop rectangle focused on the main object.
	// If detection fails or no object found, it returns the input rect or a center rect.
	Detect(img *vips.ImageRef) (*image.Rectangle, error)
}

ObjectDetector defines the interface for object detection logic.