What a Borehole Image Log Is — and What Those Sinusoids Tell You

If you have ever opened a borehole image log and wondered why the geologist keeps drawing dark wavy lines across it, this piece is for you. Those waves are not decoration and they are not noise. Each one is a planar geological feature — a bedding plane, a fracture, a fault — intersecting a cylinder and being flattened onto a page. The shape it leaves behind is a sinusoid, and the height and offset of that sinusoid carry the two numbers every structural model in the subsurface depends on: dip and azimuth. Understanding why is the difference between treating an image log as a picture and treating it as a measurement.

What a borehole image tool actually does

A high-resolution borehole image logA wireline logging tool that presses arrays of microresistivity electrodes against the borehole wall to produce a high-resolution electrical image of the rock. comes from a wireline tool that does something deceptively simple: it measures the electrical conductivity of the rock at the borehole wall, point by point, all the way up the well. The instrument carries arrays of tiny electrodes on pads and flaps that press against the wall. As the tool is pulled uphole, each electrode draws a current into the formation, and the rock's resistance to that current becomes a pixel. Conductive rock reads dark; resistive rock reads bright.

The first-generation device of this kind was introduced commercially in 1991 as the successor to earlier microscanner tools. A classic high-resolution imager carries 192 electrodes arranged in two rows, with 24 copper electrodes per pad or flap, and the pad array wraps around roughly 80% of the borehole circumference. The vertical resolution is on the order of 0.2 inches, with a sampling interval near 0.1 inches — fine enough to resolve a hairline fracture a few millimetres wide.

There is a sibling tool worth naming, because real datasets mix them. A compact microresistivity toolA slimmer through-tubing microresistivity imager with a much shallower depth of investigation than the standard high-resolution tool, used where the larger tool will not fit. trades coverage and depth of investigation for a slim body that fits where the larger imager cannot. The standard high-resolution imager reads several inches into the formation; the compact tool reads under an inch. For an interpreter, and for any model trained on both, that difference in physics matters — but the geometry we are about to describe is identical for either tool.

Unrolling a cylinder into a rectangle

Here is the conceptual move that makes everything else click. The borehole is a cylinder. The image log is a rectangle. To get from one to the other, the processing software takes a vertical cut down the borehole wall — conventionally at true north for a vertical well — and unrolls the cylinder flat, like peeling a label off a tin and laying it on the table.

The horizontal axis of the resulting image is therefore not a distance. It is an angle: azimuth around the borehole, running 0° to 360° from left to right, north back to north. The vertical axis is depth. Because the pads cover only about 80% of the wall, the unrolled image has blank vertical stripes — the inter-pad gaps where no electrode touched the rock. Those gaps appear whenever the borehole circumference exceeds the pad coverage, and filling them sensibly is its own problem. The first thing any honest interpretation pipeline does is decide how to treat that missing data rather than pretend it was measured.

Now drop a flat geological surface — say a bedding plane — through that cylinder. If the bed were perfectly horizontal, the cut would be a flat ring, and on the unrolled image it would draw a straight horizontal line. But beds are almost never perfectly horizontal. Tilt the plane, and the line it traces around the cylinder rises on one side and falls on the other. Unroll that, and the rising-and-falling line becomes a wave that completes exactly one full cycle across the width of the image. That wave is a sine curve. Not approximately — exactly, as a matter of geometry. A tilted plane intersecting a cylinder always produces an ellipse, and an ellipse unrolled against an angular axis is a sinusoid.

Reading dip out of amplitude

Once you accept that every planar feature is a sinusoid, the rest is bookkeeping. In practice an interpreter — or a curve-fitting routine — fits each feature with a function of the form

y = A · sin(ω·x + φ) + offset

where x is the azimuth in degrees. The constant that converts those degrees into radians for a single full cycle is 0.0175 (that is π/180), so the working form a fitting script actually uses looks like A · sin(0.0175·x + φ) + offset. Four numbers fall out, and each one means something physical.

The amplitude A is the half-height of the wave — how far the crest rises above the trough. That height is set entirely by how steeply the plane is tilted relative to the borehole. A gentle dip barely lifts the wave off the horizontal; a steep dip throws a tall sinusoid across the whole image. For a vertical well of known diameter, the relationship is direct: the steeper the dip, the taller the sinusoid. This is why, when geoscientists move from vertical to horizontal wells, the same beds suddenly produce much shorter sinusoids — the geometry between the borehole and the bedding has changed, and fracture-detection models built for vertical wells have to shrink their analysis window accordingly (in one carbonate play, from an 800-pixel patch down to roughly 200 pixels) just to keep the smaller waves in frame.

The offset is just where the feature sits in depth — the centre line of the wave. And the phase φ is the horizontal shift of the sinusoid: it tells you where around the borehole the wave reaches its lowest point. That trough points in the down-dip direction, which is exactly the dip azimuth. So phase encodes azimuth the same way amplitude encodes dip. Two of the four fitted parameters are the answer the structural geologist came for.

Apparent dip, true dip, and the depth budget

Two complications keep this honest, and both matter enormously the moment you try to automate it.

First: the dip and azimuth you read directly off the unrolled image are apparent, measured relative to the borehole axis. To get true dip and azimuth — relative to the earth — you have to correct for how the well itself is deviated and oriented in space. That correction needs the tool-angle and well-angle channels, which ride along in the same binary wireline log fileThe binary wireline log file that carries logging-tool measurements and metadata, including the orientation channels needed to convert apparent dip to true dip. as the image. When those orientation channels are missing, you are stuck with apparent values, and any model forced to fall back from true to apparent dip degrades visibly. The geometry is clean; the metadata is where reality intrudes.

Second: resolution sets a hard floor on precision. In one Middle East carbonate dataset, a single pixel of the unrolled binary log image corresponded to about 3 cm of depth. That means a built-in ±3 cm depth uncertainty is present in every pick, before any interpreter or model does anything, simply because you cannot locate a feature more precisely than one pixel. It is a useful number to carry in your head: when someone reports a depth-matching accuracy, ask what the pixel budget was, because no method beats its own raster.

Why this is the foundation under the AI

Everything above is why borehole-image machine learning looks the way it does. A model that detects fractures and beds is not classifying pictures; it is recovering the parameters of sinusoids. The most effective approaches treat each feature as an object to be detected and then regress its depth, dip, and azimuth directly — predicting the sinusoid parameters end to end rather than first painting a pixel mask and fitting curves afterward. On the 14-well Middle East carbonate dataset behind much of this work, that direct-regression approach reached roughly 75% fracture F1 at a 5 cm depth threshold, with dip accuracy near 90% at a 3° tolerance and azimuth accuracy near 92% for fractures at a 15° tolerance against expert picks.

The colour convention is part of the signal, too. On a borehole image log, darker means more conductive, which usually means an open fracture filled with conductive drilling mud; brighter means resistive, which points to a sealed or mineralised feature — calcite, anhydrite, or a healed hairline. Interpreters formalise this into a small taxonomy: continuous versus discontinuous, conductive (dark) versus resistive (bright). A model that ignores brightness throws away the distinction between a fracture that flows and a fracture that is cemented shut — which is most of the economic question in a fractured reservoir.

So the next time you see those dark waves marching down an image log, read them the way the tool wrote them: a flat plane, cut by a cylinder, unrolled onto a page, leaving a sinusoid whose height is a dip and whose shift is a compass bearing. Get that geometry into your bones and every downstream choice — how to fill the pad gaps, why pixel size caps your precision, what a transformer is actually predicting — stops being arbitrary and starts being obvious.

References

[1] Service-company fullbore microresistivity imager — tool description and microresistivity imaging principles. https://www.slb.com/

[2] Energistics, Digital Log Interchange Standard (RP66) specification for binary wireline log files. https://www.energistics.org/