Enum sn0int_std::gfx::HashAlg
source · pub enum HashAlg {
Mean,
Gradient,
VertGradient,
DoubleGradient,
Blockhash,
Median,
// some variants omitted
}
Expand description
Hash algorithms implemented by this crate.
Implemented primarily based on the high-level descriptions on the blog Hacker Factor written by Dr. Neal Krawetz: http://www.hackerfactor.com/
Note that hash_width
and hash_height
in these docs refer to the parameters of
HasherConfig::hash_size()
.
Choosing an Algorithm
Each algorithm has different performance characteristics
Variants§
Mean
The Mean hashing algorithm.
The image is converted to grayscale, scaled down to hash_width x hash_height
,
the mean pixel value is taken, and then the hash bits are generated by comparing
the pixels of the descaled image to the mean.
This is the most basic hash algorithm supported, resistant only to changes in resolution, aspect ratio, and overall brightness.
Further Reading: http://www.hackerfactor.com/blog/?/archives/432-Looks-Like-It.html
Gradient
The Gradient hashing algorithm.
The image is converted to grayscale, scaled down to (hash_width + 1) x hash_height
,
and then in row-major order the pixels are compared with each other, setting bits
in the hash for each comparison. The extra pixel is needed to have hash_width
comparisons
per row.
This hash algorithm is as fast or faster than Mean (because it only traverses the hash data once) and is more resistant to changes than Mean.
Further Reading: http://www.hackerfactor.com/blog/index.php?/archives/529-Kind-of-Like-That.html
VertGradient
The Vertical-Gradient hashing algorithm.
Equivalent to Gradient
but operating on the columns of the image
instead of the rows.
DoubleGradient
The Double-Gradient hashing algorithm.
An advanced version of Gradient
;
resizes the grayscaled image to (width / 2 + 1) x (height / 2 + 1)
and compares columns
in addition to rows.
This algorithm is slightly slower than Gradient
(resizing the image dwarfs
the hash time in most cases) but the extra comparison direction may improve results (though
you might want to consider increasing
hash_size
to accommodate the extra comparisons).
Blockhash
The Blockhash.io algorithm.
Compared to the other algorithms, this does not require any preprocessing steps and so may be significantly faster at the cost of some resilience.
The algorithm is described in a high level here: https://github.com/commonsmachinery/blockhash-rfc/blob/master/main.md
Median
The Median hashing algorithm.
This is a variation of the Mean hashing algorithm that uses the median instead of the mean for filtering. This variation is discussed in the further reading section, and an implementation can be found here https://github.com/JohannesBuchner/imagehash/blob/4de9becdb13ecad67b7393cc17b5a44ea1c61b6b/imagehash.py#L193:
Further Reading: http://www.hackerfactor.com/blog/?/archives/432-Looks-Like-It.html
Trait Implementations§
source§impl<'de> Deserialize<'de> for HashAlg
impl<'de> Deserialize<'de> for HashAlg
source§fn deserialize<__D>(
__deserializer: __D
) -> Result<HashAlg, <__D as Deserializer<'de>>::Error>where
__D: Deserializer<'de>,
fn deserialize<__D>( __deserializer: __D ) -> Result<HashAlg, <__D as Deserializer<'de>>::Error>where __D: Deserializer<'de>,
source§impl PartialEq<HashAlg> for HashAlg
impl PartialEq<HashAlg> for HashAlg
source§impl Serialize for HashAlg
impl Serialize for HashAlg
source§fn serialize<__S>(
&self,
__serializer: __S
) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>where
__S: Serializer,
fn serialize<__S>( &self, __serializer: __S ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>where __S: Serializer,
impl Copy for HashAlg
impl Eq for HashAlg
impl StructuralEq for HashAlg
impl StructuralPartialEq for HashAlg
Auto Trait Implementations§
impl RefUnwindSafe for HashAlg
impl Send for HashAlg
impl Sync for HashAlg
impl Unpin for HashAlg
impl UnwindSafe for HashAlg
Blanket Implementations§
§impl<'a, T> AsTaggedExplicit<'a> for Twhere
T: 'a,
impl<'a, T> AsTaggedExplicit<'a> for Twhere T: 'a,
§impl<'a, T> AsTaggedImplicit<'a> for Twhere
T: 'a,
impl<'a, T> AsTaggedImplicit<'a> for Twhere T: 'a,
source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere T: ?Sized,
source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
source§impl<Q, K> Equivalent<K> for Qwhere
Q: Eq + ?Sized,
K: Borrow<Q> + ?Sized,
impl<Q, K> Equivalent<K> for Qwhere Q: Eq + ?Sized, K: Borrow<Q> + ?Sized,
source§fn equivalent(&self, key: &K) -> bool
fn equivalent(&self, key: &K) -> bool
key
and return true
if they are equal.