Skip to content

Operator Tiers

ACL Pack operators are grouped into three tiers based on your license. Each higher tier includes all operators from the previous tier.

Tier Overview

TierOperator familiesPlansDescription
Core57Trial (1 sample), Starter, Pro, Business, EnterpriseEssential image processing — filters, color conversion, geometric transforms, arithmetic, analysis, drawing
Advanced100Pro, Business, EnterpriseAdds DFT/FFT, HSV/Lab color, advanced filters, feature detection, contour analysis, histogram analysis
Full113Business, EnterpriseComplete standard library — Business YUV remap, distance analysis, full feature suite

Trial limits to 1 sample operator: resize, at fixed 1920×1280 resolution with a watermark on the output.

Standard image pixel types are uint8_t, uint16_t, and float. Exact type support is operator- and backend-dependent; auxiliary types such as int16_t, int64_t, and double are used only in specific output, accumulator, geometry, or parameter roles.


Filter

OperatorCoreAdvFullNEON
gaussianBlur
boxFilter
medianFilter
morphology (erode/dilate)
sobel3x3
canny
laplacian
stackBlur
filter2D
sepFilter2D
scharr
bilateralFilter
guidedFilter
unsharpMask
gaborFilter
nlMeansDenoising
edgePreservingFilter
detailEnhance
mergeMertens
tonemap

Color Conversion

OperatorCoreAdvFullNEON
RGB2Gray
channelSwap
rgb2YUV_fixed
yuv2RGB_fixed
rgb2YUV_float
yuv2RGB_float
yuvSwing
gammaTransform
bayerCvtColor
rgb2HSV
rgb2Lab

Geometric

OperatorCoreAdvFullNEON
resize [trial]
rotate
pyrDown
pyrUp
buildPyramid
resizeYUV✓ (NEON-only)
rotateYUV✓ (NEON-only)
YUV_utilities (Enterprise custom)✗ (CPP-only fused pipeline)
YUV_utilities_float (Enterprise custom)✗ (CPP-only fused pipeline)

resizeYUV / rotateYUV are NEON-only atomic YUV-format operations (NV21/NV12/YV12/YUV444). YUV_utilities / YUV_utilities_float are CPP-only fused pipelines (resize + rotate + YUV→RGB in one pass) delivered as Enterprise customization outside the standard Core / Advanced / Full package counts.

Arithmetic

OperatorCoreAdvFullNEON
addImg
absDiff
mul
bitwise (AND/XOR/NOT)
lut
threshold
addWeighted
convertScaleAbs
inRange
normalize
alphaImgFusion
adaptiveThreshold
phaseMagnitude
linearTransform2x2

Analysis

OperatorCoreAdvFullNEON
histogram
integral
copyMakeBorder
equalizeHist
minMaxLoc
blockAverage
mean
count
clahe
histMatch
matchTemplate
moments
findContours
extractBlockPixels
connectedComponent_8n_dfs
connectedComponentLabeling
distanceTransform
distanceTransformU8

Feature Detection

OperatorCoreAdvFullNEON
harris
shiTomasi
fast
shiTomasiDetect
orb
sift
surf
hog
houghLines
houghLinesP
houghCircles
opticalFlowLK
bfMatch
bfMatchBinary
bfKnnMatch
bfKnnMatchBinary

Transform

OperatorCoreAdvFullNEON
warpAffine
warpPerspective
remap
getAffineTransform
getRotationMatrix2D
getPerspectiveTransform
findHomography
dltHomography
homographyError
yuvRemap

Math

OperatorCoreAdvFullNEON
dft1d
dft2d
dftReal1d
idftReal1d
mulSpectrums

Drawing

OperatorCoreAdvFullNEON
drawLine
drawRect
drawCircle
putText

Contour Analysis

OperatorCoreAdvFullNEON
contourArea
arcLength
boundingRect
convexHull
approxPolyDP
minAreaRect
fitEllipse

Memory

OperatorCoreAdvFullNEON
cropRect

Tier Totals

CategoryCoreAdvFull
Filter111620
Color Conversion91111
Geometric777
Arithmetic111314
Analysis81318
Feature Detection01416
Transform6910
Math055
Drawing444
Contour Analysis077
Memory111
Subtotal57100113

Counting convention: The 113 category rows above count each standard documented operator family once (CPP and NEON variants of the same operator share one row). Enterprise-custom YUV_utilities* fused pipelines are listed for visibility but are not part of the standard package counts. Counting customer-facing callable entries — including N-image variants, kernel-size dispatchers, and CPP/NEON pairs that ship as separate callable functions — the Business package exposes 163 callable entries. Tier licensing is per-operator family — all variants of a Starter-tier operator are included in the Starter tier. The [trial] marker indicates the operator is the sample operator included in the Trial edition.