Skip to content

Config

Config dataclass 

Config(gamma: float = 1.0, gray_weights: tuple[int, int, int] = (38, 75, 15), phash_dct_size: int = 32, phash_reduction_size: int = 8, radial_projections: int = 180, radial_samples: int = 128, radial_sigma: float = 3.5, bmh_block_size: int = 16, mhash_alpha: float = 2.0, mhash_level: float = 1.0, mhash_size: int = 512, whash_mode: WHashMode = 'fast', whash_remove_max_haar_ll: bool = False, grayscale: bool = False, alpha_mode: AlphaMode = 'grey', auto_orient: bool = True, max_pixels: int = 256 * 1024 * 1024, decode_scale: DecodeScale = 1)

How images are loaded and hashed: every setting of a libphash context.

The defaults are libphash's. A Config is checked when it is created, against the ranges the C library accepts, so a bad value fails here with the name of the field and not at the first hash. It is immutable; :meth:replace gives a modified copy. The same object configures an :class:ImageContext, :func:hash and a batch.

Hash values depend on the settings: two hashes are comparable only when computed with equal configurations (and the same libphash version).

gamma class-attribute instance-attribute 

gamma: float = 1.0

Gamma applied before the Radial hash's projections; read by Radial only.

gray_weights class-attribute instance-attribute 

gray_weights: tuple[int, int, int] = (38, 75, 15)

Relative R, G, B weights of the grayscale conversion.

phash_dct_size class-attribute instance-attribute 

phash_dct_size: int = 32

pHash: side of the DCT grid (phash_reduction_size..32).

phash_reduction_size class-attribute instance-attribute 

phash_reduction_size: int = 8

pHash: side of the low-frequency block the hash is made of, 4..8 (the hash has its square in bits).

radial_projections class-attribute instance-attribute 

radial_projections: int = 180

Radial: number of lines through the center, 40..4096.

radial_samples class-attribute instance-attribute 

radial_samples: int = 128

Radial: points read on each line, 2..4096.

radial_sigma class-attribute instance-attribute 

radial_sigma: float = 3.5

Radial: Gaussian blur sigma before the projections, (0, 64/3].

bmh_block_size class-attribute instance-attribute 

bmh_block_size: int = 16

BMH: side of the block grid, 2..32 (the digest has its square in bits).

mhash_alpha class-attribute instance-attribute 

mhash_alpha: float = 2.0

mHash: scale base (> 1).

mhash_level class-attribute instance-attribute 

mhash_level: float = 1.0

mHash: scale exponent (>= 0).

mhash_size class-attribute instance-attribute 

mhash_size: int = 512

mHash: normalization side, 62..4096.

whash_mode class-attribute instance-attribute 

whash_mode: WHashMode = 'fast'

wHash: "fast" (16x16, one Haar level) or "full" (ImageHash's scale).

whash_remove_max_haar_ll class-attribute instance-attribute 

whash_remove_max_haar_ll: bool = False

wHash: zero the coarsest LL coefficient (ImageHash's remove_max_haar_ll).

grayscale class-attribute instance-attribute 

grayscale: bool = False

Decode to one gray channel: faster, but the color algorithms then raise RequiresColorError.

alpha_mode class-attribute instance-attribute 

alpha_mode: AlphaMode = 'grey'

How an alpha channel is resolved: blended over grey, white or black, or ignored.

auto_orient class-attribute instance-attribute 

auto_orient: bool = True

Apply the EXIF orientation of JPEG and PNG files and buffers.

max_pixels class-attribute instance-attribute 

max_pixels: int = 256 * 1024 * 1024

Largest image accepted, in pixels; 0 leaves only the library's fixed limit.

decode_scale class-attribute instance-attribute 

decode_scale: DecodeScale = 1

Decode at 1/1, ½, ¼ or ⅛ of the resolution where the decoder can (JPEG).

replace 

replace(**changes: Any) -> Config

Return a copy with the given fields changed, checked like a new Config.